Relative Observer and Technical Error in the Ultrasonic Monitoring of Thyroid Nodular Growth Using Nodular Volume (VOL), vs. Longest Dimension (LD), vs. the Sum of the 3 Nodular Dimensions (SUM3D)

Abstract

Abstract Though not a reliable indicator of malignancy, ultrasonic monitoring of nodule growth still has a role in the evaluation of nodules, e.g. indicating when a nodule may require biopsy or re-biopsy. Observer and technical limitations, however, limit the precision of ultrasonic determination of simple growth, vs. stability or shrinkage. Ultrasonic parameters used for this purpose all have their own limitations. Monitoring nodule growth by VOL frequently exhibits wide and conflicting swings in apparent size compared to the penultimate size, doubtlessly reflecting measurement limitations. As a growth parameter, LD typically exhibits a smoother time course but does not address growth in the other two dimensions. SUM3D includes changes in all dimensions but, like LD, is not a true measure of nodule mass or volume. This study was to determine the relative error of these three growth parameters and how it relates to their relative efficacy for nodular growth monitoring. The anterior-posterior (AP), left-right (LR) and superior-inferior (SI) dimensions of 34 benign nodules were determined ultrasonographically by four pairs of trained observers. One observer of a pair was regarded as a Time-1 observer and the other as a Time-2 observer, simulating the process for determining growth change over time. All observers measured the same image of each of the 34 nodules but were unaware of the measurements obtained by any other observer. For each image for each pair of observers, the dimensions were used to calculate the VOL, LD and SUM3D and the perceived changes thereof from Time-1 to Time-2. Since only one image for each nodule was distributed, differences between the Time-1 vs Time-2 measurements for each nodule could only reflect observer-based differences. “S”-curves plotting the nominal %-change in a parameter reported by the Time-2 observer compared to that reported by the Time-1 observer (x-axis) were rank-ordered from negative to positive changes (y-axis). The %-change in each parameter due to observer/technical error ranging from the highest over-estimate to the lowest were, in order from Top 10%, Middle 40%, and Bottom 10%, respectively: LD: 19-36%, -4 to 6%, -15 to -42%; SUM3D: 15 to 28%, -4 to 4%, -11 to -43%; VOL: 48 to 105%, -13 to 15%, -33 to -81%. The magnitude of %-change from Time-1 to Time-2 for the VOL parameter were 2 to 3 times greater than that of the LD or SUM3D parameters for the top 10% of values, the middle 40% of values, and the bottom 10% of values. These degrees of difference coincide with the wide variability seen in nodular growth curves [not illustrated here] when nodular VOL (y-axis) is plotted as a function of length of observation (x-axis). This study helps explain why monitoring nodular growth by LD or by the SUM3D usually provides a clearer, less fluctuant illustration of thyroid nodule growth over time than does VOL.