This study sought to establish the reproducibility of volumetric software in measuring small non-metastatic pulmonary nodules < 150 mm3, using established methods. Given the effects of partial voluming on CT, and the non-spherical nature of many nodules detected incidentally in the clinical and lung screening setting, the limits of agreement were expected to be wider than published in previous similar studies involving metastatic nodules. Unexpectedly, this study demonstrated narrower limits of agreement, 15.5% either side of the mean relative difference in nodule volume. The upper and lower limits of agreement for the relative difference in nodule volumes measured by two observers were only ± 4%, confirming data from previous studies that interobserver variability in nodule volume measurements contributes very little towards the overall variability in nodule measurements [9, 10].
Compared to previous similar studies where the morphological characteristics of nodules have been compared [8, 10], a greater proportion of the nodules in the current study were irregular in contour, as was predicted due to the non-metastatic nature of nodules included in this study. Despite this, the current study has shown narrower limits of agreement for the relative nodule volume measurements than the majority of previous studies [8,9,10, 15, 16]. This may reflect advances in volumetry software over the past 15 years, differences in volumetry software packages, and advances in CT scanner technology. One study by Zhao et al demonstrated narrower 95% limits of agreement (− 12.1 to + 13.4%) in the volumetric measurements of a cohort of 32 known non-small-cell lung cancers; however, the mean tumour size in this study was > 3-cm diameter, and results are therefore not directly comparable to the current study of very small lung nodules . The impact of nodule size on measurement variability has been studied in a small number of previous in vivo studies with conflicting results [8,9,10, 15]. Wormanns et al found very similar limits of agreement in nodules ≤ 10 mm in diameter, compared to those > 10 mm in diameter , whereas Goodman et al found that confidence limits narrow with increasing nodule volume . Comparing the volumetric measurements of metastatic nodules < 8 mm and ≥ 8 mm in diameter, De Hoop et al found that the interscan variability in nodule volumes decreased with increasing nodule volume when measured by 5 of 6 nodule volumetry software packages . By contrast, Talwar and colleagues found lower variability for nodule measurements in nodules < 500 mm3 compared to those over 500 mm3 . However, no previous studies have focussed specifically on small nodules < 150 mm3 (equivalent to < 6.59-mm diameter), frequently encountered in both the lung cancer screening setting, and incidentally on CT scans of the chest. This study demonstrated no evidence that the relative difference in nodule volumes (%) varies in relation to the absolute size of the nodule for small nodules between 30 and 150 mm3.
A number of nodule management guidelines in the clinical and screening setting recommend a 3- or 6-month follow-up CT for solid nodules in the size range 30–150 mm3 [5, 7, 19, 20], the lower limit applying to new nodules developing on incident round screening CTs . Results from this study indicate that > 15% growth in a nodule volume may represent true nodule growth in this cohort of small nodules, and that an assumption of stability should not be made for nodules growing 15–25% over 3–6 months. This assumption of stability may result in a short delay to investigation of a cohort of nodules measuring 115–150 mm3, and these may require closer short-term surveillance, for example, with a repeat scan in a further 3–6 months, particularly as nodules approach a threshold for intervention (> 200 mm3). This would prevent a potential delay to diagnosis of malignancy in growing nodules in this size range. Provided that a minimum volume threshold of 200 mm3 is maintained with nodule management protocols prior to intervention, little harm is expected to result from lowering the threshold to determine growth to 15% in this cohort of nodules 115–150 mm3, since at least one further short-term interval scan would be required to confirm persistent growth prior to any invasive procedure. This study did not evaluate nodules in the size range 150–300 mm3 which are also considered indeterminate in a number of nodule management protocols utilised in screening [5, 18,19,20, 24]. As several previous studies have incorporated larger nodules and found wider limits of agreement (around ± 25%), maintaining the threshold of 25% growth to confirm true growth may be appropriate for larger nodules to prevent potential over-investigation of nodules in the range 150–300 mm3. Further studies are warranted in this regard.
Strengths of the study include that all individuals were scanned with an identical low-dose scanning protocol on the same scanner, and analysis was performed using a modern volumetry software package. Therefore, many of the conditions in which this study was performed are likely to closely mirror conditions in a lung screening cohort. A limitation of the current study is that the results have been obtained using a single software package and therefore may not apply to other volumetry packages. Studies of similar patient cohorts using other software packages are warranted. It is noted that, in clinical practice, further variability may be introduced through several other variables including CT acquisition factors (such as dose), reconstruction techniques (filtered back projection or iterative reconstruction), and reconstruction parameters including slice thickness. Of these factors, reconstruction algorithm and slice thickness are the primary contributors to interscan variability . As much as possible, all such parameters must be kept constant between scans comparing nodule volumes. In the event that scan acquisition and reconstruction parameters are not constant between scans, it may be more appropriate to use the 25% threshold for growth currently in use. However, in the screening setting, scans are ordinarily performed with identical scanning parameters and therefore the findings of our current study would apply in this context. It is also noted that the findings of this study should not be extrapolated to nodules which are poorly segmented by volumetry software and therefore measured by diameter. Such nodules may require ongoing surveillance for an extended period (up to 2 years) based on nodule diameter.
In conclusion, this study has demonstrated that, for small non-metastatic pulmonary nodules, true growth can be reliably concluded to have occurred with a volume change of > 15% where scanning parameters are identical between scans. Caution, therefore, should be exercised in participants with nodules growing 15–25%, particularly those nodules in the range of 115–150 mm3. Under current nodule management guidelines widely used in the clinical and screening setting, such nodules would be presumed to be stable, and closer surveillance of these nodules may be warranted.