Table 3

Linear regression analysis using the best independent correlates of BE from each domain

Correlation coefficientp Value
Best independent correlates from each domain
 Visceral adipose tissue volume by DXA scan+0.500.001
Lung function
 FEV1%, supine−0.40<0.001
Sleep variables
 Mean overnight oxygen saturation−0.50<0.001
Ventilatory control
 Ventilatory response to 15% oxygen−0.280.02
Respiratory muscle strength
 Sniff maximum (cm H2O)−0.280.02
Hormonal, nutritional and inflammatory measures
 Vitamin D−0.300.01
Cumulative correlation coefficientp Value
Significant independent correlates
 Visceral adipose tissue volume by DXA scan+0.490.004
 Ventilatory response to 15% oxygen+0.580.04
  • Results for the best correlates of BE in a linear regression for each individual domain. The correlate was chosen on the basis of the likelihood of it having a causal association with BE (eg, ventilatory response to 15% oxygen was chosen over the absolute fall in SaO2, and vitamin D was chosen over both BNP and transferrin). The lower part of the table shows the result for the best overall significant independent correlates of BE in a multiple linear regression containing the best correlate from each domain. A cumulative correlation coefficient of 0.58 indicates that the model accounts for 34% of the variance in BE. Vitamin D levels were the next most significant correlate, but with a p value of 0.19.

  • BE, base excess; BNP, brain natriuretic peptide; DXA, dual-energy X-ray absorptiometry; FEV1, forced expiratory volume in 1 s.