Fig 5. NBS linear regression: Predicted adolescent lung function is more strongly associated with early childhood FEV₁ than with percentage of cultures positive for Pa in early childhood.

Subject values for mean early childhood and adolescent FEV₁ are shown in scatterplot, with subjects Pa positive in early childhood represented by gray triangles, Pa negative by black circles. Superimposed lines represent NBS linear regression results showing predicted adolescent FEV₁ (with 95% confidence intervals—dashed lines) by early childhood FEV₁. The black line shows predicted adolescent FEV₁ by early childhood FEV₁ when percentage of cultures positive for Pa prior to age 6 is held at 0% (25^th percentile); the grey line represents the predicted adolescent FEV₁ when percentage of cultures positive for Pa prior to age 6 is held at 30% (approximately the 90^th percentile for our cohort). These values were all adjusted for gender, year of birth, age at diagnosis, number of cultures before age 6, BMI Z score at 2 to 4 years of age, and number of PFT tests included in early childhood and adolescent values (all held at mean). Predicted mean adolescent FEV₁ increased by 19.7 percent predicted when early childhood FEV₁ was increased from the 25^th percentile (85.5% predicted) to the 95^th percentile (117.4%), while it decreased by only 4.8% when percentage of cultures positive for Pa was increased from the 25^th percentile (0%) to the 92^nd percentile (30% positive), or 5.8% when increased to 36.8% positive (95^th percentile). Contribution to the explained variance of the full model was 0.1957 for mean early childhood FEV₁ vs. 0.0157 for percentage of cultures positive for Pa before age 6.