Table 2 Summary of statistical methods
From: A focused review of statistical practices for relating radiation dose-volume exposure and toxicity
Characteristic | N (%) of 325 |
|---|---|
Analytic approach1 | |
 Non-directional test of means | 102 (31.3) |
 Logistic regression | 180 (55.4) |
 Linear regression | 39 (12.0) |
 Cox proportional hazards | 67 (20.6) |
 Other | 75 (23.1) |
No. of OARs assessed | |
 1 | 194 (59.7) |
 2 | 55 (16.9) |
   ≥ 3 | 76 (23.3) |
No. of DVH parameters assessed | |
 1–5 | 109 (33.5) |
 6–20 | 141 (43.4) |
 21–100 | 54 (16.6) |
  > 100 | 18 (5.5) |
 Not explicitly stated | 3 (0.9) |
Method of selecting DVH parameters | |
 Prior research | 28 (11.7) |
 Institutional treatment protocol | 18 (5.5) |
 Not stated | 279 (85.8) |
Multivariable statistics | |
 Yes | 226 (69.5) |
 No | 99 (30.5) |
Multiple comparisons performed | |
 Yes | 258 (79.4) |
 No | 67 (20.6) |
Correction for multiple comparisons2 | |
 None | 214 (82.9) |
 Bonferroni | 15 (5.8) |
 Bootstrapping | 12 (3.7) |
 Benjamini-Hochberg | 5 (1.5) |
 Cross validation | 3 (0.9) |
 Other | 9 (2.8) |
Cut points used | |
 Yes | 154 (47.4) |
 No | 171 (52.6) |
Method of determining cut points3 | |
 Prior research or protocol | 8 (5.2) |
 Set to mean, median, or quartile values | 17 (11.0) |
 Maximize discriminant value from ROC curve | 63 (40.9) |
 Recursive partitioning | 9 (5.8) |
 Not stated | 23 (14.9) |
 Other | 37 (24.0) |
Non-dosimetric variables included | |
 Yes | 242 (74.5) |
 No | 83 (25.5) |