Receiver operating characteristic (ROC) analysis

A ROC curve is a plot of the true positive rate against the false positive rate for the different possible cut-off points of a diagnostic test. Specificity and sensitivity are the measures used to determine the diagnostic accuracy of a test (14). Using SPSS (version 24; IBM Corp.), ROC curves were constructed through a comparison of the individual systematic errors with the various MPTV values at the three different levels (upper neck, lower neck and head). The combined test summarized the diagnostic accuracy of a test by means of a single number. The curve is associated with numerous advantages including the illustration of all of the cut-off points of a diagnostic test. It also reveals the associations between the sensitivity of a test and its specificity, though it is not affected by the prevalence of a condition (for example in the present study, the prevalence of setup errors) and it is possible for researchers to compute important summary measures of accuracy using these curves (14).

For example, if one wished to test if a certain sample has a condition under investigation, in the present study this being the presence of setup errors in patients undergoing RT, the case sensitivity would be the proportion of true positives (those with actual setup shifts). The specificity on the other hand is the proportion of true negatives-the proportion of the cases that had no setup errors among those that did not have any shifts.

In a ROC curve the sensitivity (true positive rate) is plotted against the false positive rate (1-Specificity) for different cut-off points. Any point on the ROC plot resembles a sensitivity corresponding to a particular decision threshold. A test that has perfect discrimination (the absence of an overlap between two distributions), will produce a curve that passes through the upper left corner (100% sensitivity, 100% specificity). Hence, the closer the ROC plot is to the upper left corner, the greater the overall accuracy of the test (14).