Variation in tumor burden is the current reference standard for assessment of response to chemotherapy.1 Phase II clinical trials reported the association between tumor shrinkage and improvement in overall survival (OS).2 Furthermore, randomized phase III trials showed inverse correlation between evolution of tumor size and time to event.3 Currently, computed tomography (CT) imaging is the elective tool for assessment of tumor size to test the effect of chemotherapy.
Since 2000, standardized CT assessment of tumor burden has been performed by the Response Evaluation Criteria in Solid Tumors (RECIST).4 RECIST criteria rely upon variation in axial longest diameter (LD) of tumor lesions on CT.4 In 2009, the RECIST criteria were revised, redefining the maximum number of target lesion to be assessed, the management of lymph node evaluation, and the interpretation of newer imaging technology results, such as positron emission tomography (PET) and magnetic resonance.1 These criteria are widely adopted for evaluation of tumor response, and regulatory authorities have accepted them as appropriate guidelines.1
However, RECIST criteria are prone to several limitations: unidimensional measurement suffers from interobserver and intraobserver variability5,6; in addition, changes in the axial LD inconstantly reflect the actual evolution of tumor, and its prognostic significance is still unclear in advanced lung cancer.7–10 Therefore, RECIST’s limitations might bias the assessment of patient outcome and lead to inappropriate changes in therapy.11 Indeed, oncologists are advocating more accurate imaging systems to evaluate response after therapy either in clinical practice or in drug trials.10
Commercially available semiautomatic software programs currently provide both volume and diameters of tumors, based on segmentation algorithms. It has been shown that volumetric measure provides a more comprehensive estimate of tumor burden.12 Recent studies showed higher reproducibility for the semiautomatic measurements of lung cancer compared with manual diameter; however, they comprised a relatively small number of lesions.5,6,13 Furthermore, the response to therapy as assessed by the volume changes of advanced lung cancer showed controversial relationships with survival.9,14 Therefore, the reproducibility and prognostic significance of unidimensional and volumetric measurements in lung cancer are still under debate.
The purpose of this study was to compare the reproducibility of semiautomatic and manual CT measurements of 94 lung lesions in patients with locally advanced or metastatic non–small cell lung cancer (NSCLC). Later, we aimed to test whether early variation in tumor size was predictive of patient OS.
MATERIALS AND METHODS
The study protocol was approved by our Institutional Review Board. Informed consent was waived for the retrospective examination of both clinical and radiologic data. Patients were retrospectively searched for in the electronic oncology records of the Medical Oncology Unit of our hospital from January 2007 to December 2012. Clinical and radiologic records were jointly reviewed by an oncologist (M.T., with 8 y of experience) and a radiologist (N.S., with 10 y of experience). Inclusion criteria were as follows: (a) pathology-proven nonresectable NSCLC in stages III to IV; (b) homogenous platinum-based chemotherapy; (c) and contrast-enhanced CT scan of the chest at the time of diagnosis (baseline) and after at least 1 cycle of chemotherapy (follow-up). According to RECIST criteria, patients who exhibited nonmeasurable pulmonary lesions such as solid nodules with axial LD <10 mm were excluded from the study.1
A total of 47 consecutive patients (27 male and 20 female; median age 67 y, range 36 to 80 y) were included: 16/47 (34%) patients had stage III NSCLC, and 31/47 (66%) patients had stage IV NSCLC. Seventeen of 47 (36%) patients had extrathoracic disease.
Entire chest CT scans were obtained either with a 64-slice system (Somatom Sensation 64; Siemens Medical Solutions, Forchheim, Germany) or with a 128×2-slice system (Somatom Definition Flash; Siemens Medical Solutions) with spiral cranio-caudal acquisition in full inspiration, with the patient in the supine position. The acquisition parameters used for the 64–detector-row scanner were as follows: median tube current 152 mAs (range, 75 to 378 mAs), tube voltage 120 kV, and pitch 1.5. Acquisition parameters for the 128×2–detector-row scanner were as follows: median tube current 230 mAs (range, 101 to 399 mAs), tube voltage 120 kV, and pitch 1.5. In each chest CT scan the tube current was modulated by automatic exposure control (Care Dose; Siemens Medical Solutions). Single-phase acquisition was obtained after 35 seconds of peripheral intravenous power injection of 90 to 120 mL of nonionic contrast material with a 300 mg/mL iodine concentration (Iomeron 300; Bracco, Milano, Italia) at 3 mL/s according to patient size. Image data sets were reconstructed using a medium-soft tissue kernel (B30) at a slice thickness setting ranging from 1.5 to 2.0 mm.
All CT scans were anonymized and transferred to a dedicated workstation (Syngo MultiModality workplace, version VE40A; Siemens Healthcare, Forchheim, Germany) that included the Syngo CT Oncology Software (Siemens Healthcare). This software is commercially available and was approved by the Food and Drug Administration. The application provides a range of automated tools specifically designed to support physicians in the segmentation and volumetric evaluation of suspicious lesions, including dedicated tools for the lung. Nevertheless, as the lung nodule algorithm is optimized for small structures (<2 cm), it was not applicable in our study, which assessed larger pulmonary lesions. Thus, we used the generic nonspecific segmentation algorithm instead of the lung nodule algorithm.
Reader 1 (N.S., a radiologist with 10 y of experience in interpreting chest CT scans) selected the largest measurable pulmonary lesion (according to RECIST guidelines) for each patient, classifying it as follows: (a) “central” tumors when the lesion showed contact with the hilar structures; and (b) “peripheral” tumors when the center of the lesion was within the parenchyma and had no contact with hilar structures—that is, within the lung parenchyma, with or without contact with the subcostal pleura.4 Thoracic nodes and extrathoracic lesions were not measured. The pulmonary lesions were independently evaluated by 2 radiologists, reader 1 (N.S.) and reader 2 (I.M., with 8 y of experience in interpreting chest CT scans); 2 trainees, reader 3 and reader 4 (C.M. and D.C., each with 1 y of experience in interpreting chest CT scans), further measured all lesions selected. Thus, the largest pulmonary lesion of each patient was measured by 4 readers with different levels of experience.
Baseline and follow-up CT scans of each patient were displayed simultaneously on a dedicated workstation. The images were reviewed using the standard “mediastinum” window settings (window width=350; window center=50); however, each reader had the opportunity to change the window settings to better assess contours of the lesion. The 4 readers independently measured the axial LD (M-LD) of each preselected lesion with a digital manual caliper. Thereafter, the readers ran the computer-aided evaluation (as detailed below) (Figs. 1A, B). Each observer was blinded to the other readers’ measures. Reader 3 repeated manual and semiautomatic measurements 10 months later to test intraobserver agreement.
Semiautomatic Software Analysis
The segmentation consisted of drawing a rough diameter across the lesion in 1 of the displayed planes (axial, coronal, or sagittal). The segmentation algorithm estimated thresholds by density histogram analysis around the lesion. Preliminary segmentation was generated by means of the 3-dimesional (3D) region-growing technique and estimated thresholds. Adjacent structures of similar density were separated by morphologic operations. Final segmentation of the lesions is displayed in 3-orthogonal views and in 3D volume-rendering reconstruction. The reader verified the lesion segmentation, and, in case of missegmentation, the entire process was restarted by repositioning another diameter and/or by editing edges with the dedicated semiautomatic corrector tool (Figs. 2A, B). Each operator detailed any segmentation correction. Finally, the software calculated the volume and axial LD (SA-LD) of each lesion. The total time required to perform the semiautomatic lesion assessment was recorded for each reader.
Response to therapy for each lesion was based upon unidimensional measurements following RECIST 1.1 guidelines.1 Patients with a resolving lesion were defined as achieving complete response (CR); subjects with a decrease of >30% in the LD of the lesion were defined as achieving partial response (PR); subjects with an increase >20% of LD (of 5 mm at least) were defined as having progressive disease (PD).1 Volumetric response was based upon the thresholds suggested by the RECIST 1.0 criteria and obtained from linear cutoffs assuming spherical shape of the lesion.4 A volumetric decrease >65% was deemed PR, and a volumetric increase >73% was deemed PD.4 Lesions with size changes out of the considered thresholds were classified as stable disease (SD). Patients who achieved a CR or PR were defined as “radiologic responders,” whereas patients who demonstrated SD or PD were defined as “radiologic nonresponders.”10 Thus, patients with a volumetric decrease >65% or a reduction in axial LD >30% were considered RECIST-based radiologic responders. In addition, we arbitrarily evaluated the thresholds for the detection of responders and nonresponders by considering the quartiles of the variation in M-LD, SA-LD, and volume between baseline and follow-up CTs. Furthermore, the response of nodal thoracic metastases and appearance of new extrathoracic lesions at the time of follow-up CT were recorded by reviewing both clinical and radiologic (integrated fluorodeoxyglucose [FDG] PET/CT results) data.
The Wilcoxon test was used to compare M-LD and SA-LD. The difference in size between central and peripheral lesions was tested with the Mann-Whitney U test. The concordance correlation coefficient (CCC) was initially calculated to quantify intrareader and interreader agreement for the manual and semiautomatic measurements.15 A CCC value of +1 indicated perfect agreement, whereas −1 indicated perfect reverse agreement. The following descriptive scale was suggested for the interpretation of the CCC: <0.90, poor; 0.90 to 0.95, moderate; 0.95 to 0.99, substantial; and >0.99, almost perfect.16 Furthermore, the variation coefficient percentage (VC%) was calculated as the ratio of the standard deviation to the mean to test measurement reproducibility among distributions with different mean values.17 The weighted κ coefficient of agreement (κw) was used to quantify interreader and intrareader variation for categories of radiologic responders and nonresponders as obtained following both RECIST criteria and experimental thresholds applied to manual and semiautomatic measurements; the agreement between readers was classified as poor (κ=0.00 to 0.20), fair (κ=0.21 to 0.40), moderate (κ=0.41 to 0.60), good (κ=0.61 to 0.80), and very good (κ=0.81 to 1.00).18
The OS was calculated from the date of the initial CT to the time of death from any cause. Demographics, staging, and CT-based categories were correlated with OS using the Kaplan-Meier method (product-limit). The survival functions were compared between independent groups of patients by means of the log-rank test. Cox proportional hazards regression analysis was used to examine the association between prognostic variables and OS to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). The relative changes in manual and semiautomatic measurements (as continuous variables), RECIST categories (assessed through manual and semiautomatic measurements—RECIST M-LD, RECIST SA-LD, RECIST VOL), response categories based on the volume thresholds (ΔVOL), thoracic nodal response, and the occurrence of new extrathoracic lesions at the time of follow-up CT were tested as potential predictors in the Cox model. Significant variables at univariate analysis were included in the multivariate analysis using a stepwise regression model. A P value <0.05 was considered significant. Statistical analysis was performed with SPSS software version 20 (SPSS Inc., Chicago, IL) and MedCalc version 220.127.116.11 (Broekstraat, Mariakerke, Belgium).
The demographic characteristics of the study patients are summarized in Table 1. Each reader evaluated 47 lesions both at baseline and at follow-up CT (total n=94). All lesions were solid. The first follow-up CT was performed after a median interval of 2 months (range, 1 to 9 mo) from the baseline examination; however, excluding 2 patients with the follow-up CT obtained after 7 and 9 months, the maximum time interval from baseline CT was 4 months. Eighteen of 47 (38%) lesions were centrally located, whereas 29/47 (62%) lesions were peripheral. Software missegmentation was recorded in 20/94 (21%) lesions by reader 1, in 14/94 (15%) lesions by reader 2, in 11/94 (12%) lesions by reader 3, and in 5/94 (5%) lesions by reader 4. In 3 of 4 readers, the software missegmentation was mostly observed in central lesions: missegmentation of central lesions occurred in 7/20 (35%) cases for reader 1, in 10/14 (71%) cases for reader 2, in 9/11 (81%) cases for reader 3, and in 4/5 (80%) cases for reader 4. Each reader recorded a maximum of 2 correction attempts. The readers obtained satisfactory segmentation results (including the correction procedure) in a median time of 3 minutes (range, 2 to 8 min).
Data on tumor size and relative percentage changes over time are given in supplemental material (Tables S1 and S2, Supplemental Digital Content 1, http://links.lww.com/JTI/A56). The manual diameter was smaller than the corresponding semiautomatic diameter both at baseline and at follow-up CT (P<0.001), although the percentage changes over time of the same measurements were similar (P=0.24). This was true regardless of the lesion’s location. No significant difference was found between central and peripheral lesions regarding both size and percentage variation as obtained manually and semiautomatically (P=0.13 to 0.55).
Intrareader and Interreader Agreement
Table 2 shows the interreader and intrareader agreement for both manual and semiautomatic measurements. The interreader agreement of manual and semiautomatic measurements in overall lesions was substantial (Table 2). Reproducibility of manual diameter was moderate in central lesions and substantial in peripheral lesions, whereas semiautomatic measurements showed moderate interreader agreement in lesions centrally located and almost perfect reproducibility in peripheral lesions (Table 2). However, there was greater interreader agreement for the assessment of both volume (CCC: 0.974 to 0.991; VC%: 5.6% to 9.5%) and semiautomatic diameter (CCC: 0.980 to 0.987; VC%: 6% to 7.3%) as compared with manual diameter (CCC: 0.950 to 0.984; VC%: 6.4% to 11.7%) of the lesions. Overall, interreader agreement was greater for peripheral lesions. In particular, the manual diameter of peripheral lesions (CCC: 0.968 to 0.991; VC%: 5.5% to 10.6%) demonstrated higher reproducibility in comparison with the central one (CCC: 0.899 to 0.967; VC%: 7.3% to 12.9%). Furthermore, volume (CCC: 0.977 to 0.996; VC%: 4.2% to 10.4%) and semiautomatic diameter (CCC: 0.990 to 0.994; VC%: 4.5% to 5.9%) were more reproducible in peripheral lesions compared with volume (CCC: 0.961 to 0.984; VC%: 5.5% ot 8.3%) and semiautomatic diameter (CCC: 0.945 to 0.979; VC%: 6.1% to 9.7%) in central findings. There was similar intrareader agreement for manual diameter (CCC: 0.984; VC%: 6.4%), semiautomatic diameter (CCC: 0.986; VC%: 6.2%), and volume (CCC: 0.990; VC%: 5.8%). The intrareader agreement of volume and semiautomatic diameter was greater for peripheral lesions (Table 2).
By categorizing patients as responders or nonresponders, the reader agreement was good to very good for both central and peripheral lesions (Table 2). The best intrareader agreement for distinguishing responders from nonresponders was found by using the 70% volume decrease cutoff (κw=0.98 to 1.00), as compared with RECIST thresholds of manual diameter (κw=0.61 to 0.83), semiautomatic diameter (κw=0.84 to 0.87), and volume (κw=0.64 to 0.88).
Analysis of Survival
The median observation time from baseline CT was 40 months (range, 15 to 83 mo). Thirty-eight of 47 (77.1%) patients died during this time interval, with an estimated median OS of 14 months (95% CI for the median: 8.8-19.2 mo). Among demographics and baseline tumor stage parameters, only N3 stage was significantly associated with shorter OS (HR: 5.3; 95% CI: 1.1-27.5; P<0.05).
Patients with new extrathoracic metastases showed significantly (P<0.001) shorter OS (median OS: 5 mo; 95% CI for the median: 3.7-6.3 mo) compared with the remaining patients (median OS: 17 mo; 95% CI for the median: 12.4-21.6 mo). The Cox models confirmed the association between the occurrence of new extrathoracic metastases and poorer survival (HR: 5.6 to 9.4; lower 95% CI: 2.3-2.9; upper 95% CI: 13.4-29.7; P<0.001) for all readers. The association between tumor size variation and OS was investigated for each reader by plotting Kaplan-Meier curves of patient category according to either RECIST thresholds or the quartiles of proportional size changes at follow-up CT. Manual and semiautomatic RECIST thresholds did not stratify subjects with different survival durations in all readers (P=0.06 to 0.66) (Table 3; Figs. 3A–C). Among the unidimensional and volumetric quartiles of proportional size changes, only a decrease in volume ≤70% was associated with shorter OS for 3 out of 4 readers (P<0.05) (Table 3; Fig. 3D). For the same readers, patients with decrease ≤70% were associated with shorter OS in the Cox model both at univariate and at multivariate analyses, as shown in Table 4 (HR: 5 to 22.2; lower 95% CI: 1.2-2.5; upper 95% CI: 20.8-201.1; P<0.05).
This study shows high reproducibility of both manual and semiautomatic measurements of advanced NSCLC lesion size on CT. However, higher interreader agreement was observed for semiautomatic measurements, compared with the M-LD. The early response assessment based on RECIST criteria either by unidimensional or by 3D evaluation of lesion size did not predict survival, whereas a decrease >70% of the initial tumor volume was associated with longer OS.
The higher reproducibility of the semiautomatic measurements in comparison with the manual assessment of tumor dimensions in patients suffering from advanced NSCLC is in keeping with prior studies.6,13 Dinkel et al6 reported higher reproducibility of semiautomatic measurement compared with our results. This difference could be due to several factors, including the smaller prevalence of central lesions, the lower number of lesions assessed, and the higher rate of segmentation correction.6 Likewise, Zhao and colleagues reported slightly higher intrareader agreement. Several reasons could explain this difference, notably the lesion location, the performance of the volume analysis software (homegrown vs. commercially available), and the higher rate of edited segmentation.5 By contrast, the lower interreader and intrareader VCs% found in another study might be expected because of the inclusion of a higher number of smaller lesions that show higher variability.19 Nishino et al13 suggested that the tumor location did not influence the variability of its measurements. Nonetheless, the mediastinal infiltration was not taken into account.13 In the present study, lower intrareader and interreader agreement of both the manual and the semiautomatic assessment was found for central lesions. Therefore, contiguity with structures of similar density (ie, mediastinum) and ill-defined borders seen in central tumors make the dimensional evaluation more challenging and less accurate.
In keeping with prior studies, RECIST criteria did not predict survival in advanced NSCLC or, at least, predicted survival only when PD was compared with other RECIST classes in a larger cohort of patients.7–10 It has been suggested that volumetric measurement could more accurately represent biological tumor burden and could be regarded as a promising tool to assess treatment response.20 In contrast, Knollmann et al9 found that volumetric RECIST classification of the response was not associated with longer survival in patients suffering from advanced NSCLC. For a deeper understanding of this lack in clinical results, it is worth mentioning that volumetric RECIST thresholds (≥65% decrease for PR, ≥73% increase for PD) were geometrically calculated from linear cutoffs assuming spherical shape of the lesion.4 Notably, pulmonary lesions typically appear with irregular shape and margins (eg, lobulation and spiculation) that could hardly be simplified with geometrical sphere. A recent study reported that response assessment based on calculated tumor volume as obtained from the unidimensional measurement using sphere equation was only moderately concordant with that achieved by the semiautomatic volume in advanced NSCLC patients.21 Therefore, new thresholds for volumetric response assessment are required. Nishino et al14 reported that patients suffering from epidermal growth factor receptor (EGFR) mutant-NSCLC with early volume decrease higher than 38% had longer survival compared with patients with lesions with a more limited decrease in size. In our study, an upward trend of the volumetric decrease threshold (70%) allowed the distinction of patients with longer survival in a different population for both histology (non–EGFR-mutant NSCLC vs. EGFR-mutant NSCLC) and chemotherapy (platinum-based chemotherapy vs. erlotinib or gefitinib) relative to the above-mentioned study.14 Notably, volumetric changes may be dependent on several factors such as tumor type and therapy and are variably associated with OS.22,23 Furthermore, the heterogenous composition of tumors (eg, neoplastic cells, stromal tissue, and inflammatory cells) has been shown to influence radiologic assessment.8,24 Fibrosis, necrosis, and abundant inflammatory infiltrate of up to 50% of the tumor mass were demonstrated in resected NSCLC after induction of platinum-based chemotherapy.25 Thus, tumor size changes observed at CT might be confounded by inflammatory or fibrotic components. However, we speculate that a higher threshold could be more specific in the assessment of actual reduction of viable tumor cells.
The present study has several limitations. First, the significant association demonstrated between shorter survival and patients with volume decrease ≤70% is limited by the small sample size; nevertheless, this association is strengthened by its reproducibility among 3/4 readers with different levels of experience in thoracic imaging. Second, the study included only the largest pulmonary lesion measurement for each patient. Measuring a single lesion is the usual approach in studies focused on response assessment of NSCLC using advanced imaging techniques.26,27 Furthermore, several authors reported that the reduction in target lesion number did not affect response assessment in NSCLC and other tumors; notably, in one series the same treatment response classification was obtained by measuring both the single largest lesion and up to 5 lesions in >90% of metastatic colon cancer patients.28 Furthermore, changes in mediastinal nodal sizes at CT were not added to the LD of the pulmonary lesions. However, CT alone showed lower accuracy in detecting nodal response in comparison with integrated FDG-PET/CT (60% vs. 83%).29 Thus, nodal response was assessed by reviewing integrated FDG-PET/CT data records and was considered as single predictor of survival. Third, only the first 2 CT examinations were evaluated; nonetheless, response to treatment is assumed to be greater early, and only the initial tumor changes would indicate nonresponders in time, when a change in therapy makes most sense.9 Fourth, the interval between baseline and follow-up CT was wide with a maximum period of 9 months in 1 patient; nevertheless, follow-up CT was homogenously obtained for all patients after the first cycle of chemotherapy, and the vast majority of the study patients were rescanned after 3 to 4 months.
In conclusion, CT assessment of lesion size obtained by commercially available semiautomatic software showed better reproducibility in comparison with manual unidimensional evaluation in advanced NSCLC. In addition, RECIST criteria for early response assessment by unidimensional or 3D measurements failed to classify patients with better survival. In contrast, a decrease of >70% in lesion volume identified patients with longer survival. Further investigations, on a larger population, are warranted to validate this volumetric threshold for response assessment to better guide therapeutic decisions in advanced NSCLC.
1. Eisenhauer EA, Therasse P, Bogaerts J, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer. 2009;45:228–247
2. Paesmans M, Sculier JP, Libert P, et al. Response to chemotherapy has predictive value for further survival
of patients with advanced non-small cell lung cancer: 10 years experience of the European Lung Cancer Working Party. Eur J Cancer. 1997;33:2326–2332
3. Buyse M, Thirion P, Carlson RW, et al. Relation between tumor response to first-line chemotherapy and survival
in advanced colorectal cancer: a meta-analysis. Meta-Analysis Group in Cancer. Lancet. 2000;356:373–378
4. Therasse P, Arbuck SG, Eisenhauer EA, et al. New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada. J Natl Cancer Inst. 2000;92:205–216
5. Zhao B, James LP, Moskowitz CS, et al. Evaluating variability in tumor measurements from same-day repeat CT scans of patients with non-small cell lung cancer. Radiology. 2009;252:263–272
6. Dinkel J, Khalilzadeh O, Hintze C, et al. Inter-observer reproducibility of semi-automatic tumor diameter measurement and volumetric analysis in patients with lung cancer. Lung Cancer. 2013;82:76–82
7. Lara PN, Redman MW, Kelly K, et al. Disease control rate at 8 weeks predicts clinical benefit in advanced non-small-cell lung cancer: results from Southwest Oncology Group randomized trials. J Clin Oncol. 2008;26:463–467
8. Birchard KR, Hoang JK, Herndon JE Jr, et al. Early changes in tumor size in patients treated for advanced stage non-small cell lung cancer do not correlate with survival
. Cancer. 2009;15:581–586
9. Knollmann FD, Kumthekar R, Fetzer D, et al. Assessing response to treatment in non-small-cell lung cancer: role of tumor volume
evaluated by computed tomography. Clin Lung Cancer. 2014;15:103–109
10. William WN Jr, Pataer A, Kalhor N, et al. Computed tomography RECIST assessment of histopathologic response and prediction of survival
in patients with resectable non-small-cell lung cancer after neoadjuvant chemotherapy. J Thorac Oncol. 2013;8:222–228
11. Erasmus JJ, Gladish GW, Broemeling L, et al. Inter-observer and intra-observer variability in measurement of non-small-cell carcinoma lung lesions: implications for assessment of tumor response. J Clin Oncol. 2003;21:2574–2582
12. Heussel CP, Meier S, Wittelsberger S, et al. Follow-up CT measurement of liver malignoma according to RECIST and WHO vs. volumetry. Rofo. 2007;179:958–964
13. Nishino M, Guo M, Jackman DM, et al. CT tumor volume
measurement in advanced non-small-cell lung cancer: performance characteristics of an emerging clinical tool. Acad Radiol. 2011;18:54–62
14. Nishino M, Dahlberg SE, Cardarella S, et al. Tumor volume
decrease at 8 weeks is associated with longer survival
in EGFR-mutant advanced non-small-cell lung cancer patients treated with EGFR TKI. J Thorac Oncol. 2013;8:1059–1068
15. Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics. 1989;45:255–268
17. Jones R, Payne BJones R, Payne B. Analytical methods: control and comparison. Clinical Investigation and Statistics in Laboratory Medicine. 1997 London ACB Venture Publications:41–42
18. Brennan P, Silman A. Statistical methods for assessing observer variability in clinical measures. BMJ. 1992;304:1491–1494
19. Zhao B, Tan Y, Bell DJ, et al. Exploring intra- and inter-reader variability in uni-dimensional, bi-dimensional, and volumetric measurements of solid tumors on CT scans reconstructed at different slice intervals. Eur J Radiol. 2013;82:959–968
20. Bru A, Albertos S, Subiza JL, et al. The universal dynamics of tumor growth. Biophys J. 2003;85:2948–2961
21. Nishino M, Jackman DM, DiPiro PJ, et al. Revisiting the relationship between tumour volume and diameter in advanced NSCLC patients: an exercise to maximize the utility of each measure to assess response to therapy. Clin Radiol. 2014;69:841–848
22. Nougaret S, Rouanet P, Molinari N, et al. MR volumetric measurement of low rectal cancer helps predict tumor response and outcome after combined chemotherapy and radiation therapy. Radiology. 2012;263:409–418
23. Liu F, Zhao B, Krug LM, et al. Assessment of therapy responses and prediction of survival
in malignant pleural mesothelioma through computer-aided volumetric measurement on computed tomography scans. J Thorac Oncol. 2010;5:879–884
24. Choi H, Charnsangavej C, Faria SC, et al. Correlation of computed tomography and positron emission tomography in patients with metastatic gastrointestinal stromal tumor treated at a single institution with imatinib mesylate: proposal of new computed tomography response criteria. J Clin Oncol. 2007;25:1753–1759
25. Soussan M, Cyrta J, Poliquen C, et al. Fluorine 18 fluorodeoxyglucose PET/CT volume-based indices in locally advanced non-small cell lung cancer: prediction of residual viable tumor after induction chemotherapy. Radiology. 2014;272:875–884
26. Fraioli F, Anzidei M, Zaccagna F, et al. Whole-tumor perfusion CT in patients with advanced lung adenocarcinoma treated with conventional and antiangiogenetic chemotherapy: initial experience. Radiology. 2011;259:574–582
27. Yabuuchi H, Hatakenaka M, Takayama K, et al. Non-small cell lung cancer: detection of early response to chemotherapy by using contrast-enhanced dynamic and diffusion-weighted MR imaging. Radiology. 2011;261:598–604
28. Zacharia TT, Saini S, Halpern EF, et al. CT of colon cancer metastases to the liver using modified RECIST criteria: determining the ideal number of target lesions to measure. Am J Roentgenol. 2006;186:1067–1070
29. De Leyn P, Stroobants S, De Wever W, et al. Prospective comparative study of integrated positron emission tomography-computed tomography scan compared with remediastinoscopy in the assessment of residual mediastinal lymph node disease after induction chemotherapy for mediastinoscopy-proven stage IIIA-N2 non-small-cell lung cancer: a Leuven Lung Cancer Group Study. J Clin Oncol. 2006;24:3333–3339