Table 3.
Bias, Precision and Accuracy in Validation for Different Models for Estimating GFR
| Model | Difference | Percent Difference | ||||
|---|---|---|---|---|---|---|
| Median | IQR** | Median | IQR* | P30 | RMSE | |
| (95% CI)* | (95% CI) | (95% CI) | (95% CI) | (95% CI) | (95% CI) | |
| Internal Validation (n=1045) | ||||||
| 1. Cystatin C | 0 (−1,1) | 13 (12,14) | 0 (−2,2) | 29 (27,31) | 82 (81,83) | 0.257 (0.238, 0.275) |
| 2. Cystatin C, age, sex and race | 0 (0,1) | 12 (11,13) | 1(−1,3) | 28 (26,29) | 84 (83,85) | 0.250 (0.231, 0.267) |
| 3. Creatinine, age, sex, and race | 0 (−1,1) | 10 (9,11) | 1 (−1,2) | 26 (24,28) | 85 (84,86) | 0.226 (0.209, 0.244) |
| 4. Cysatin C and creatinine, age, sex and race | 0 (0,1) | 10 (9,11) | 0 (−1,2) | 23 (21,24) | 89 (88,90) | 0.200 (0.184, 0.215) |
| 5. MDRD Study equation | 1 (0,1) | 10 (9,11) | 2 (1,3) | 26 (24,28) | 85 (84,86) | 0.226 (0.210, 0.244) |
| 6. Average #1 & # 5 eGFR Scys & MDRD | 1 (0,1) | 10 (9,11) | 2 (0,3) | 23 (21,24) | 88 (87,89) | 0.204 (0.190, 0.220) |
| External Validation (n=438) | ||||||
| 1. Cystatin C | −3 (−3,−2) | 8 (7, 9) | −10 (−13,−7) | 31 (28, 36) | 73 (72,74) | 0.264 (0.239, 0.289) |
| 2. Cystatin C, age, sex and race | −2 (−2,−1) | 8 (7, 9) | −6 (−8,−4) | 30 (26, 32) | 79 (78,80) | 0.248 (0.223, 0.271) |
| 3. Creatinine, age, sex, and race | 2 (1,3) | 8 (7, 9) | 7 (4,9) | 25 (22, 29) | 84 (83,85) | 0.229 (0.210, 0.247) |
| 4. Cysatin C and creatinine, age, sex and race | 0 (0,1) | 7 (6, 8) | 1 (−1,3) | 22 (20, 25) | 90 (89,91) | 0.193 (0.174, 0.211) |
| 5. MDRD Study equation | 2 (2,3) | 8 (7, 9) | 8 (6,11) | 24 (22, 27) | 85 (84,86) | 0.231 (0.213, 0.249) |
| 6. Average #1 & # 5 eGFR Scys & MDRD | 0 (−1,1) | 7 (6, 8) | 1 (−2,3) | 24 (22, 27) | 90 (89,91) | 0.196 (0.176, 0.217) |
Difference is calculated as measured GFR-estimated GFR. Units are in mL/min per 1.73 m2. To convert GFR from mL/min/1.73 m2 to mL/sec per 1.73 m2, multiply by 0.01666. Percent difference is calculated as (measured GFR-estimated GFR)/measured GFR, and units are in percent. Median values measure bias and the inter-quartile ranges measure precision. P30 is calculated as the percentage of estimated GFR within 30% of measured GFR. IQR, inter-quartile range; RMSE, root mean square error; CI, confidence interval; MDRD Study, Modification of Diet in Renal Disease Study; eGFR, estimated glomerular filtration rate; Scys, serum cystatin C.
The 95% confidence interval around the estimates of bias, IQR, P30 and RMSE provides the range of values which is likely to include the parameter in 95% of circumstances. Comparison of confidence intervals around any metric for two equations provides information as to difference between them. If the two metrics were the same, the chance that the upper and lower bounds of the confidence intervals would not overlap is less than 5%. Therefore, it can be concluded that they are statistically significantly different.
IQR is the width of the 25th to 75th percentile. For example, the 25th to 75th of the difference is −5.4 and 7.2, for an IQR of 13. For the percent difference, 25th to 75th is −13.8% and 14.8% which leads to the calculated IQR of 29%.