224 Chapter 15 relevant correlation between invasion size ratio and mean variation in total tumor size measurement was seen (1st and 2nd round <0.001 and 0.11, respectively). This indicates that for the pathologists who assigned an invasive measurement, variation is relatively low when the invasive size is close to total tumor size, but that there is much greater variation between individuals when perceived invasion is more focal. Secondly, as the invasive line only provides information about pathologists scoring “invasion present” cases, we subsequently examined the relationship between the fraction of pathologists scoring a case invasive and the CVinv (Figure 9). In cases where most pathologists score “invasion present”, CVinv is low. In contrast, in cases where most pathologists favour “no invasion”, the remaining pathologists tend to assign high variation in size of invasiveness. Thus, there is less consensus in the assessment of invasive areas in cases which most pathologists regard as being wholly in situ. Figure 9. Relation between fraction of pathologists scoring “no-invasion” and the variation in invasive measurement is shown for rounds 1 and 2. Note that for cases where all 42 pathologists scored “Invasion yes” (fraction = 0) the variation in invasive measurement is relatively low, while for a higher number of pathologists scoring “invasion no” the CV-inv for the remaining pathologists is high. Thirdly, heatmap analyses of cases with low CVinv frequently show a well-defined round hotspot area, as opposed to cases with high CVinv, where the hotspot areas are sometimes angular and assignments are in different locations (Figure 7). Thus, disagreement in the degree of invasiveness (high CVinv) is linked to disagreement about its location. All three observations point to limited accuracy and precision in the assignment of invasion in some small pulmonary adenocarcinomas. Subsequently, we assessed
RkJQdWJsaXNoZXIy MTk4NDMw