Click on image to view larger version.

Figure 2 Smoothed five-year predicted probability of developing atherosclerotic cardiovascular disease by level of glomerular filtration rate (GFR). Unadjusted curve shows the risk incorporating each individual’s value for 19 covariates in Table 3. Adjusted curve shows the average risk in the Atherosclerosis Risk in Communities study population if everyone had the GFR value on the X-axis. The linear model includes GFR as a continuous variable in a Cox regression while the cubic spline includes a cubic transition between linear segments with knots (at 0.05, 0.275, 0.5, 0.725, 0.95 quantiles of GFR) corresponding to GFR values of 63.7, 81.7, 88.2, 101.1, and 123.7 ml/min/1.73 m², respectively. Tick marks along the X-axis indicate GFR values for individual participants (the marks form a solid bar in GFR regions with many individuals). A lower GFR cut-off of 30 ml/min/1.73 m² was chosen because only 12 subjects had GFR values between 15 and 30 ml/min/1.73 m²; therefore, the data was less precise in the latter range.