[Table A posted as supplied by author]

�

Table A Clinical characteristics

Measure	Males	Females
Age (years)	41(15)	40 (13)
Count	798	1948
Obese/non-obese	440 / 358	1026 / 922
Body mass index, kg/m2	32.4(9.4)	31.6(9.3)
Body fat, %	29.6(16.0)	44.4(17.3)
Waist circumference, cm	109.3(25.5)	97.9(21.6)
Waist-to-hip ratio	0.987(0.336)	0.871(0.098)
Systolic blood pressure, mmHg	134.5(19.7)	124.9(18.3)
Diastolic blood pressure, mmHg	81.1(13.5)	77.6(22.2)
HOMA, index	5.18(8.10)	3.29(4.34)
P-Insulin, mU/L	17.37(19.11)	12.34(10.55)
P-Glucose, mmol/L	5.87(1.96)	5.45(1.84)
HDL-Cholesterol, mmol/L	1,16(0.34)	1.38(0.39)
P-Triglycerides, mmol/L	1.88(1.69)	1.36(0.85)
S-Leptin� m g/L	15.96(14.75)	28.87(23.05)

All values are mean with SD within brackets. Obesity is defined as a body mass index> 30 kg/m2

�

Legend to web figure [posted as supplied by author]

Upper panel A: Relationship between HOMA-index and waist circumference. HOMA index data was plotted in logarithmic scale and curves were fitted by linear regression analysis. The horizontal line indicates cut-off for HOMA-index and the vertical line cut-off for waist circumference, dividing the plot in a 2 by 2 field table, visualizing true and false cases. Lower panel B: Receiver operator characteristics curve analyses of the waist circumference accuracy to detect insulin resistance. Curves were fitted by non-linear regression analyses. Open circles denote females and closed circles denote males.

[Further statistical details posted as supplied by author]

Further statistical details

The principal study sample is the 2746 subjects who have been recruited consecutively at our unit between 1997-2003 in order to study factors influencing body weight. The primary multivariate regression analysis was performed with HOMA-index as the dependent variable on 2314 complete cases. 432 cases were excluded due to missing values in at least one variable (age, leptin, systolic blood pressure, log P-triglycerides, HDL-cholesterol, BMI, waist-hip-ratio, waist circumference, % body fat). After removal of the non-significant covariates from the model (age, leptin, waist-hip-ratio, % body fat) the number of complete cases rose to 2423 and a new regression analysis was performed. Finally, the EM-algorithm was used to estimate missing values in the remaining 323 missing cases and the multivariate regression analysis was run in all 2746 cases showing superb agreement with the previous runs. The standardised partial regression coefficients are as follows: (waist-circumference β₁=0.36, log-plasma triglycerides β₂=0.22, systolic blood pressure partial β₃=0.09, HDL-Cholesterol β₄=-0.08 and body mass index β₅=0.18, Adjusted r² =0.573, p<0,001). We conclude that in this large sample the missing cases have no influence of the main results.