Abstract
Aim: To compare iodine-related and fluorine-18 fluorodeoxyglucose (18F-FDG) parameters during staging of lung cancer as well as during early follow-up, while investigating potential use and possible substitutability in the assessment of therapeutic response or prediction. Patients and Methods: Patients (n=45) with confirmed lung cancer underwent 18F-FDG positron-emission tomography (PET) using single-source dual-energy computed tomography was performed for staging and early follow-up. Correlation of FDG uptake and iodine-related parameters was assessed and comparison with therapy response was performed. Results: A strong correlation was found between the volumetric FDG parameters metabolic tumour volume (MTV) and total lesion glycolysis (TLG) and iodine uptake (IU) in staging (IU vs. MTV: rs=0.894; p<0.001 and IU vs. TLG: rs=0.874; p<0.001) and follow-up (IU vs. MTV: rs=0.934, p<0.001 and IU vs. TLG: rs=0.935, p<0.001). We also found significant correlation of change in these values between timepoints. We observed a significant correlation of IU, MTV and TLG with early therapy response and IU was found as a possible strong predictor. Conclusion: Strong correlation of IU and volume-based FDG parameters was proved in staging, follow-up and change during therapy. Potential role of IU in prediction of early therapy-response was identified. Our study suggests a significant benefit of using the dual-energy computed tomography as a part of 18F-FDG PET/CT in patients with lung cancer.
- Multidetector computed tomography
- positron-emission tomography
- iodine
- fluorodeoxyglucose
- lung cancer
- dual-energy
Monitoring of antitumour therapy is still mostly based on the assessment of anatomical parameters, even though this approach offers only limited informative value, particularly in the case of early response with a rather short interval from the therapy onset. The routinely used Response Evaluation Criteria in Solid Tumours (RECIST) are simple, reproducible and give clear results (1). However, using RECIST is becoming increasingly challenging in the evaluation of new therapeutic approaches. In particular, RECIST is markedly limited in early therapy response evaluation, when shorter intervals from the therapy onset affect its usability. It has been shown that the effect of new types of antitumour therapy can be observed very shortly after administration, however, the effect can be better quantified by functional rather than anatomical parameters (2). This is crucial when applying new therapeutic procedures that are highly financially demanding and thus the pressure on precise and early evaluation of response is still increasing.
Several studies have proven the ability of 18F-fluorodeoxyglucose (18F-FDG) uptake in assessing the therapy effects as well as in predicting therapeutic response (3, 4). Tumour perfusion is another important and promising functional parameter with proven added value in advanced evaluation (5, 6). Dynamic contrast-enhanced volume perfusion computed tomography (VPCT) is an established method but its inclusion in routine diagnostic and monitoring procedures is difficult due to the complicated scanning technique and image analysis. In comparison, dual-energy (DE)-CT scanning is simpler and in its image analysis, it allows easy and precise quantification of iodine-related attenuation that corresponds with tumour vascularization (2, 7, 8). DE-CT scan can also be easily integrated into positron-emission tomography (PET)/CT scanning equipment and such devices equipped with DE scanning or spectral analysis of X-rays, allowing parallel detection of metabolism and vascularization are likely to be readily available in the future.
The goal of our study was to compare iodine-related parameters and FDG parameters during staging as well as during early follow-up of the therapy effect, while assessing any possible substitutability in the assessment of therapeutic response or prediction.
Patients and Methods
Study design. The prospective study was carried out in 2016-2019 and a total of 45 patients (31 males; aged 65.8 years on average) were included. Inclusion criteria were histologically proven adenocarcinoma and advanced stage disease (IIIB and IV). All patients underwent 18F-FDG-PET/CT including single-source DE-CT scan due to staging. Early follow-up examination was then performed using the same scanning protocol after 2-3 cycles of combined chemotherapy and inhibitor of vascular endothelial growth factor (paclitaxel with carboplatin plus bevacizumab). Early therapy response was evaluated using RECIST version 1.1 (complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD)].
Patients with standard contraindications (history of severe allergic reaction and renal failure) to the iodine contrast agent were excluded from the study. The use of defined PET/CT examination protocol was approved by the Institutional Ethics Committee and all patients signed informed consent.
18F-FDG-PET/CT protocol. All examinations were performed on a clinical PET/CT scanner with integrated 128-row CT and 4-ring PET subsystem (Biograph mCT 128; Siemens Healthcare, Knoxville, TX, USA). The examination was initiated with a CT scan performed by the single-source CT subsystem following the administration of iodine contrast agent (Iomeron 350; Bracco, Milano, Italy). A uniform amount of 100 ml was injected (flow rate 4 ml/s) via the antecubital vein using an automated high-pressure injector (Ulrich Motion, Ulm, Germany) followed by a saline flush (50 ml). The first scan (single-energy) was performed of the whole body (from skull base to groins) in arterial phase identified using a dedicated monitoring system (BolusTracking; Siemens Healthineers, Forchheim, Germany). The subsequent DE scan was performed of the thorax using a prototype scanning protocol consisting of two separate scans with different fixed tube voltage, quality-referenced radiation produced, and pitch factor (80 kV/298 mAs/0.6 and 140 kV/71 mAs/1.2). The remaining scanning parameters were identical for both scans (collimation 128×0.6 mm, rotation time 0.5 s). The default delay of the single-source DE scan was set to 75 s after the end of the first, single-energy scan. Two datasets with slice thickness of 1.0 mm (increment 0.6 mm; kernel B26) were reconstructed.
The PET scan was initiated 60-70 minutes after administration of 18F-FDG (activity of 2.5 MBq/kg). The PET scan was performed by step-and-shoot technique and the whole body was covered by 5-6 scanner positions (‘beds’). The acquisition time was 90 s/bed. Raw PET data were reconstructed in a 400×400 pixel matrix in transaxial field-of-view with 46 cm diameter using the ultraHD algorithm, which combines the time-of-flight and point-spread function algorithms, and also 2 mm full width at half maximum. PET dataset with attenuation correction using single-energy scan was reconstructed for subsequent analysis.
Image analysis. PET dataset was analysed in consensus by two experienced radiologists (11 and 9 years' experience with hybrid imaging) using dedicated software application Syngo.via MM Oncology (Siemens Healthineers, Forchheim, Germany). Segmentation of the FDG uptake in the lung tumours was performed using an automatic algorithm with uniform relative isocontour threshold (42%) of absolute standard uptake value (SUV). Various SUV parameters were calculated by body weight [measured activity concentration (Bq/ml)/injected dose (Bq) × patient weight (g)]: maximum SUV (SUVmax), peak SUV (SUVpeak, highest average 1 cm3 equivalent), mean SUV (SUVmean) and metabolic tumour volume (MTV). Parameter of total lesion glycolysis (TLG) was calculated as MTV × SUVmean. Diameters in orthogonal projections and volumes were calculated in segmented lesions.
DE-CT datasets were analysed using dedicated prototype software eXamine (Siemens Healthineers) in consensus by the same experienced radiologists. Tumours were segmented using a semi-automatic algorithm with the possibility of manual corrections of peripheral borders. The value of total iodine uptake (IU; mg) and iodine concentration related to tumour size (IC; mg/ml) were acquired for all tumours.
Statistical analysis. Standard frequency tables and descriptive statistics were used to characterize the patient sample. As all investigated variables were either continuous (imaging measurements) or ordinal (therapy response), Spearman's rank correlation was used to assess their association. The predictive ability of imaging measurements for early therapy response was investigated by extending the traditional two-class receiver operating characteristic (ROC) analysis to three classes according to Nakas and Yiannoutsos (9), using an in-house written function in MATLAB (2019b; MathWorks Inc., Natick, MA, USA). For this analysis, the CR and PR categories were merged since only one case of CR was observed. Each pair of classification thresholds (one for distinguishing between CR+PR and SD, and one between SD and PD) represents a point on the ROC surface, whose coordinates are determined by the proportions of truly positively classified cases from individual classes. The further the surface arches away from the origin, i.e. the greater the volume under the surface (VUS), the better the discriminative ability of the variable is. A VUS of 1/6 (= 0.167) corresponds to random guessing (contrasting with the traditional 2D ROC area under curve, where the analogous value is 0.5), while a VUS of 1 represents perfect classification. The curves in which the ROC surface intersects the zero planes are the common 2D ROC curves for binary classification into respective category pairs.
All reported p-values are two-tailed and the level of statistical significance was set at α=0.05. Statistical processing and testing were performed in STATISTICA data analysis software system (Version 12; StatSoft, Inc., TuIsa, OK, USA).
Results
In total, 45 patients who underwent the DE-PET/CT staging examination were included in the final analysis. A follow-up DE-PET/CT was subsequently performed in 36 of them. Complete descriptive sample characteristics are summarized in Table I.
Correlation of FDG- and iodine-related parameters. A strong significant correlation was generally confirmed between IU derived from the DE-CT scan and FDG volumetric parameters (IU vs. MTV: rs=0.894, p<0.001; and IU vs. TLG: rs=0.874, p<0.001) in the staging examination. A similar strong correlation was observed in the follow-up examination (IU vs. MTV: rs=0.934, p<0.001, and IU vs. TLG: rs=0.935, p<0.001). Only moderate correlation was observed between IU and SUVmax in the staging examination (rs=0.547, p<0.001) but a strong correlation was observed in the follow-up (rs=0.785, p<0.001).
Our analysis showed strong correlation of the change in values of IU and of FDG volumetric parameters between staging and follow-up (ΔIU vs. ΔMTV: rs=0.786, p<0.001; and ΔIU vs. ΔTLG: rs=0.845, p<0.001). Only weak inverse correlations were observed between IC and volume FDG parameters. All results are summarized in Table II and Figure 1.
Prediction of therapy response. Spearman correlation analysis revealed a strong correlation between the follow-up IU and early therapy response (rs=0.710, p<0.001) and only moderate correlation between staging IU and early therapy response (rs=0.466, p<0.004). Significant correlations with response were observed for tumour volume (rs=0.463 at staging and rs=0.670 at follow-up). Moderate correlation was also found for FDG-related parameters at follow-up. Only weak or no correlation was observed between the early therapy response and IC at both staging and follow-up. Analysis of the change of parameters between the two examinations revealed strong associations with early response for ΔSUVmax and ΔSUVpeak (rs=0.670 and rs=0.607 respectively, p<0.001 for both) and weak associations for ΔSUVmean and IC. Interestingly, tumour volume and IU were no longer correlated to response in terms of their change between examinations (Table III, Figure 2).
The ability of the measured variables to predict early treatment response was also assessed using ROC analysis for a three-class classification problem (CR and PR categories were merged since only one case of CR was observed). Similarly to the Spearman correlation, the results showed moderate-quality prediction based on tumour volume or IU at staging (VUS ROC=0.544 and 0.538, respectively) and very solid discrimination for the same variables at follow-up (VUC ROC=0.740 and 0.793, respectively). Concerning the change of values between examinations, the results of the correlation analysis were confirmed again as ΔSUVmax and ΔSUVpeak showed the best ROC performance (VUS ROC=0.713 and 0.717, respectively; Figure 3).
Patient and tumour characteristics.
Discussion
FDG- and iodine-related parameters. The results of our study showed a strong correlation of IU with volume-based FDG parameters in both the staging and follow-up PET/CT examination of primary NSCLC. The correlation was slightly stronger in the case of a follow-up examination performed early after the start of neoadjuvant chemotherapy rather than that of the staging examination. The mutual relationship of vascularisation and tumour metabolism is still an often discussed and assessed topic. PET and volume-based perfusion examination are the two most frequent functional methods used for evaluation of tumorous processes but the results of studies comparing FDG and VPCT parameters to date show certain differences in assessment, which partially correspond to the varying principles of both modalities (10-13). Moreover, despite the quite long availability and proven additional value of the VPCT method in assessment of tumour tissue and prediction of therapy response, it still has not been incorporated as a routine technique in lung cancer diagnostic or monitoring procedures (14, 15).
In addition to the correlation of volume-based FDG parameters and the IU in individual examinations, our study proved a strong correlation in the change of these values between staging and early follow-up. Although this does not mean that the FDG parameters are substitutable, it suggests that the assessment of the development may offer similar informative value. On the other hand, only moderate correlation was shown between IU and standard non-volumetric FDG parameters, particularly SUVmax, which is routinely used in clinical assessment. Only weak correlation was found for IC. This fact is consistent with previous experience and the probable reason for this is the significant inhomogeneity of particularly large lung tumours, causing a bias of this parameter.
Correlation of fluorodeoxyglucose and iodine-related values.
DE-CT technology has been continuously expanding and further development such as spectral CT techniques, which allow accurate quantification of the iodine content, can be expected (16-18). Despite a long history of DE-CT technology, the number of studies comparing the iodine parameters with the metabolic activity of lung cancer lesions is still limited. Correlation of the maximum values of FDG uptake and iodine-related parameters in cases of primary lung cancer as well as for lymph nodes affected by metastasis was shown quite a long time ago by Schmid-Bindert et al. (19) However, further studies did not clearly confirm this correlation. Furthermore, varying scanning procedures as well as evaluation tools of DE-CT and FDG parameters make the correlations difficult to assess. Nevertheless, the need for the use of functional parameters not only for staging but also for more precise assessment of therapy effect is obvious (20). Recently, Ren et al. similarly demonstrated the correlation of iodine-related parameters and FDG parameters at multiple follow-up points after radiotherapy or concomitant therapy (21). Due to the aforementioned availability of PET/CT devices that allow for simultaneous DE-CT scanning, we can expect further studies on this topic, and not only in lung cancer, for which FDG-PET/CT is currently the recommended technique for staging (22). In our study, we used a scanning technique featuring DE-CT on a single-source equipment. The so-called single-source DE-CT technique provides two scans with different tube voltage (80 and 140 kV) and technical limitations (23). PET/CT scanners with fully integrated DE-CT scanning mode will eliminate these limitations.
Scatterplots showing the correlation of iodine uptake with metabolic tumour volume (MTV) and total lesion glycolysis (TLG) in staging (A and D) and follow-up (B and E) examinations, and of their change between the two examinations (C and F). Please note that all axes use a logarithmic scale.
Strong associations of the early therapy response according to the Response Evaluation Criteria in Solid Tumours (version 1.1) (1) with tumour volume and iodine uptake in staging (A and B) and follow-up examinations (C and D). When considering the change of parameters between the two examinations, change in maximum standard uptake value (ΔSUVmax) (E) and ΔSUVpeak (F) showed the strongest association with early response according RECIST. CR: Complete response; PR: partial response; SD: stable disease; PD: progressive disease.
3D Surfaces of receiver operating characteristic (ROC) analysis extended to three classes showing the capacity of tumour volume (A and C for staging and follow-up respectively), iodine uptake (B, D), change in maximum standard uptake value (ΔSUVmax) (E) and Δ SUVpeak (F) to predict/discriminate the early treatment response. The further the surface arches away from the origin, i.e. the greater the volume under the surface (VUS), the better the discriminative ability of the variable is. AUC: Area under the receiver operating characteristics curve. SD: Stable disease; PD: progressive disease.
Therapy response prediction. 18F-FDG-PET/CT was also shown to be an accurate and robust method for restaging and evaluation of therapy response (24). However, responses to selected therapies are significantly less frequently evaluated using 18F-FDG-PET/CT in common practice. The reason for this is the limited availability of the method as well as it being financially demanding and time-consuming. In addition, the RECIST criteria remain the recommended method of assessment of therapy response. These criteria are explicitly defined by the size of lesions and FDG uptake values may be used only in specific situations (25, 26). The limitations of assessing the therapy response using lesion size only are generally known but the incorporation of functional parameters in these criteria has not yet been established. CT and magnetic resonance imaging are thus still frequently used in therapy response evaluation and the information on tissue metabolism obtained during 18F-FDG-PET/CT staging cannot be used. From this perspective, a combination of 18F-FDG-PET/CT and a DE-CT scan, followed by further follow-up using DE-CT seems to be one of the possible uses of functional information during the monitoring of the therapy effect.
The extent of early response seems to be a possible predictive parameter for the overall success rate of the therapy (27). Metabolic parameters have been shown to be strong predictors of therapy response, which may be an important factor in therapy planning in the future. Satoh et al. showed that MTV and TLG values are better at predicting response than SUVmax in NSCLCs larger than 3 cm (28). Other studies confirmed volumetric values to be independent predictive factors in NSCLC, without any dependence on clinical stage (29-31). Our analysis also suggests that MTV, TLG and values at staging correlate with the early therapy effect assessed by means of the RECIST criteria, and are capable of predicting it with moderate accuracy. From this perspective, the IU value was even more significant than both of the volume-based FDG parameters.
Limitations of our study. The therapy response evaluation in our study was based on RECIST criteria, so the correlation of volume-based parameters was predictable. However, the precise determination of lesion volume from CT or MRI data is limited due to the difficult segmentation and reactive (non-tumorous) changes during therapy (e.g. atelectasis or post-irradiation fibrosis). The segmentation reliability using FDG or DE data is clearly higher than with pure CT.
Moreover, due to the association of volume-based FDG and IU parameters with the lesion volume, different diameters of tumours may lead to bias in the analysis. Our group of patients and follow-up time do not yet allow outcome analysis. Apart from the mediocre number of patients, problems include in particular the inhomogeneity of tumour differentiation and disease stage. Moreover, we assessed primary tumours only, since the combined assessment of DE-CT including metastases is currently difficult in terms of methodology. The technical limitation of the single-source DE-CT technique used was mentioned above, on the other hand though, real-time correlation with FDG parameters was performed.
Conclusion
It can be stated that our study demonstrated the strong correlation of IU in particular with the volume-based FDG parameters (MTV and TLG) during the staging or follow-up examination, as well as in the change of the values after therapy onset. In IU at staging, correlation with the early therapy response assessed using the RECIST criteria was also shown. From the perspective of future development, our study suggests a possible major benefit of using DE-CT as a part of the PET/CT examination for assessing the therapy effect and predicting therapy response.
Acknowledgements
This research was supported by the Ministry of Health of the Czech Republic, grant nr. 17-30748A.
Footnotes
Authors' Contributions
JB: Study design, clinical data and image analysis, article preparation; JL: image data analysis; MS, TF and BS: development of software for image analysis, scanning technique development and article preparation; PH: statistical analysis and article preparation; MS and MP: study design and clinical data resources; JF: study design, image analysis and article preparation. All Authors read, revised and approved the final article.
Conflicts of Interest
Thomas Flohr, Martin Sedlmair and Bernhard Schmidt: Employees of Siemens Healthineers, Forchheim, Germany. These co-authors were not involved in performing and analysing examinations.
- Received April 17, 2020.
- Revision received April 30, 2020.
- Accepted May 4, 2020.
- Copyright© 2020, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved