Neuroradiology/Head and Neck Imaging
Original Research
FDG PET/MRI Coregistration Helps Predict Response to Gamma Knife Radiosurgery in Patients With Brain Metastases
Audio Available | 
Share
OBJECTIVE. The purpose of this study was to determine whether relative standardized uptake value (SUV) measurements at FDG PET/MRI coregistration are predictive of local tumor control in patients with brain metastases treated with stereotactic radiosurgery (SRS).
MATERIALS AND METHODS. A retrospective review was conducted of the images and clinical characteristics of a cohort of patients with brain metastases from non-CNS neoplasms treated with gamma knife radiosurgery (GKRS) who underwent posttherapy FDG PET because of MRI findings concerning for progression. The PET and contrast-enhanced MR images were fused. Relative SUV measurements were calculated from ROIs placed in the area of highest FDG uptake within the enhancing lesion and in the contralateral normal-appearing white matter. Relative SUV was defined as the ratio of maximum SUV in the tumor to maximum SUV in healthy white matter. Two independent readers evaluated response to GKRS using serial posttherapy MRI performed at least 3 months after GKRS completion. The relation between relative SUV and local tumor progression was evaluated with respect to treatment effect.
RESULTS. Eighty-five patients (48 [56.5%] women, 37 [43.5%] men; mean age at diagnosis, 60.5 ± 11.3 years) met the inclusion criteria. Thirty-three (38.8%) lesions progressed after SRS. There was a significant association between relative SUV and local tumor control (p = 0.035). Relative SUV provided a diagnostic ROC AUC of 0.67 (95% CI, 0.55–0.79).
CONCLUSION. Quantitative relative SUV at posttherapy FDG PET serves as a biomarker of response to SRS in patients with brain metastases in cases in which lesion growth is identified at follow-up MRI. This prognostic data may affect management, supporting the need for further therapeutic actions for selected patients.
Keywords: FDG PET, gamma knife radiosurgery, MRI, progression, radiation necrosis
Brain metastases occur in approximately one-third of patients with a systemic cancer [1–3], a proportion that is expected to increase with improved patient survival [3]. The management of intracranial metastatic disease can markedly affect survival and quality of life. Several randomized trials and multicenter studies have shown the efficacy of stereotactic radiosurgery (SRS), whether performed alone or in combination with whole-brain radiation therapy [4–6].
SRS as a stand-alone therapy is attractive in the sense that it is noninvasive, results in minimal impact on the delivery of other treatment options, and may be synergistic with novel targeted or immunotherapeutic agents. The use of SRS alone, compared with SRS combined with whole-brain radiation therapy, results in less cognitive deterioration at 3 months in patients with brain metastases [7]. Local tumor control—or response to therapy for CNS metastasis—is achieved in most cases of metastasis, but as many as one-third of metastatic lesions exhibit initial enlargement at follow-up MRI [8]. Such a finding does not preclude response to therapy but may also represent radiation effect [8–10]. Patients who undergo immunotherapy have an increased rate of radiation necrosis after SRS [11]. As the use of immunotherapies increases, the rate of radiation necrosis may be expected to increase compared with rates in the chemotherapy era [11].
Because radiation effect is indistinguishable from true progression in a single imaging study, serial MRI is the standard for monitoring response to SRS. Radiation effect is diagnosed when the lesions become smaller or stabilize over time without further intervention [9, 10]. Despite numerous efforts with different imaging techniques [12–16], there is currently no clear consensus on a clinically available prognostic biomarker of treatment effect in a single imaging session. Such a tool would greatly help practitioners make time-efficient appropriate therapeutic decisions about the care of patients with brain metastases and lesion growth at MRI after SRS.
The objective of this study was to determine whether relative standardized uptake value (SUV) measurements at 18F-FDG PET/MRI coregistration can be used to differentiate tumor progression from treatment effect due to radiation after gamma knife radiosurgery (GKRS) in patients with brain metastases and apparent disease progression at follow-up MRI. We hypothesized that lesions with a high relative SUV compared with the normal white matter would be more likely to progress than would lesions with low relative metabolic activity.
| Materials and Methods |   |
|---|
This HIPAA-compliant study received institutional review board approval, and the requirement to obtain informed consent was waived. With use of our clinical database of GKRS patients, we retrospectively identified patients with the following inclusion criteria in a 6-year time interval: age older than 18 years, brain metastatic lesion larger than 6 mm from a known primary extracranial neoplasm, treatment with GKRS, findings of lesion growth at routine MRI after radiosurgery, FDG PET/CT of the brain after first MRI showing lesion growth, and serial follow-up MRI for at least 6 months after FDG PET. Patients retreated with GKRS after an abnormal FDG PET result were not included in the study.
Gamma knife radiosurgery—GKRS was performed with a Leksell gamma knife unit (model C, Elekta Instruments). The GKRS technique has been described previously [17]. All GKRS in this study was performed in a single session.
MRI—At our institution, patients undergo routine follow-up with MRI at least every 3 months after the completion of GKRS. At follow-up MRI, deterioration was defined as a new area of contrast enhancement or an at least 10% increase in enhancing lesion [18]. Although these findings are suspicious for tumor progression, they are not specific and may represent both progression and radiation effect. If a patient had more than one enlarging lesion, a target lesion larger than 6 mm was defined for each participant for further analysis with FDG PET coregistration and follow-up MRI. We chose 6 mm as a cutoff value to include lesions above the resolution of PET.
PET/CT—All PET examinations were performed at our institution according to a standard clinical protocol. After the patient fasted for at least 6 hours, 370 MBq of FDG was administered IV. After the patient rested for 45 minutes, PET/CT of the head was performed with an integrated system (Biograph 40, Biograph 6, or ECAT, Siemens Healthcare) For each patient, commercial software (MIMneuro 6.7, MIM Software) was used to coregister attenuation-corrected PET images to the contrast-enhanced 3D images from the most contemporaneous preceding brain MRI study.
Using gray-scale PET attenuation-corrected images and color-coded fused PET/MR images, one author (12 years of experience and fellowship training in nuclear medicine and neuroradiology) visually selected the area of greatest FDG uptake within the brain region that showed abnormal contrast enhancement at the site of the treated lesion. Circular ROIs were placed in this area and in the contralateral healthy-appearing white matter as previously described [19]. Because calculations of SUV do not completely remove variability introduced by differences in patient size, amount of injected FDG, and method of quantification [20], we used a previously validated normalized metric of metabolic activity in the enhancing tumor (relative SUV) as the ratio of maximum SUV in the tumor to maximum SUV in healthy white matter [18].
Definition of tumor progression—Response to treatment was determined by measuring the enhancing lesion on volumetric 1-mm-thick contrast-enhanced magnetization-prepared rapid-acquisition gradient-echo (MP-RAGE) T1-weighted images on serial MRI studies after FDG PET. Two board-certified fellowship-trained neuroradiologists independently evaluated for tumor progression versus adverse radiation effect. Continued increase in the size of the enhancing lesion in multiple serial examinations was considered indicative of tumor progression or treatment failure. Shrinkage or long-term stability of a previously enlarging lesion was considered to be related to radiation effect. In case of disagreement, both readers reached a consensus in a combined second reading session. Contrast enhancement was evaluated on 3D T1-weighted sagittal MP-RAGE images (slice thickness, 1 mm; matrix, 256 × 256). When available, histopathologic analysis was considered the reference standard. In lesions that were stereotactically biopsied, reader consensus was compared with pathologic result as a quality control of reader performance.
Data description—Categoric data were summarized by frequency and percentage. Continuous scaled data were summarized either by mean ± SD of the distribution or by the median of the distribution.
Analyses of predictors of tumor progression after gamma knife radiosurgery—Univariate and multivariate logistic regression were used to identify univariate predictors and multivariate predictors of outcome after GKRS. Post-GKRS tumor progression status (1, progression; 0, no progression) served as the univariate and multivariate logistic regression outcome variable. The predictor variables for tumor progression were age, sex, lesion diameter, lesion hemisphere, lesion location, GKRS tumor margin dose, time from GKRS to PET, and relative SUV. Univariate logistic regression null hypotheses were tested by Wald chi-square test. Multivariate logistic regression null hypotheses were tested by type III Wald chi-square test. A p ≤ 0.05 decision rule was used as the null hypothesis rejection criterion for both the univariate and multivariate analyses.
Agreement in definition of tumor progression—Conventional measures of diagnostic agreement (sensitivity and specificity) were used to assess the diagnostic accuracy of the readers' consensus diagnosis for tumor progression and diagnosis of tumor progression by means of pathologic analysis.
Relative standardized uptake value–related ROC analysis—As a predictor of tumor progression, ROC analysis of relative SUV was performed with the pROC package R (version 3.4, R Foundation).
| Results | |
|---|
Eighty-five patients (48 [56.5%] women, 37 [43.5%] men; mean age at diagnosis, 60.5 ± 11.3 years) met the study criteria. Of the 85 lesions, 34 (40.0%) were located in the left cerebral hemisphere, 33 (38.8%) in the right hemisphere, and 18 (21.1%) in the posterior fossa. Thirty-two of the 85 metastatic lesions (37.6%) were in frontal lobes, 15 (17.7%) in the cerebellum, 15 (17.7%) in parietal lobes, 10 (11.8%) in temporal lobes, seven (8.2%) in occipital lobes, three (3.5%) in basal ganglia, and three (3.5%) in the brainstem. The mean maximal diameter of the lesions at pre-treatment MRI was 17.4 ± 8.3 mm. Metastatic brain tumors included lung carcinoma (n = 38), melanoma (n = 17), breast carcinoma (n = 16), colorectal carcinoma (n = 3), soft tissue sarcoma (n = 2), and cardiac (n = 1), esophageal (n = 1), head and neck (n = 1), kidney (n = 1), nerve sheath (n = 1), thyroid (n = 1), ovarian (n = 1), lacrimal gland (n = 1), and neuroendocrine (n = 1) tumors.
The median interval between GKRS and PET was 180 days (95% CI, 155–205 days). Thirty-three (38.8%) lesions progressed after GKRS. There was no significant difference in demographics, lesion size and location, GKRS dose, or time from GKRS to PET between the patients whose lesions ultimately responded or progressed after therapy (Table 1).
In 17 patients (20.0%), the enhancing lesion was surgically resected and histologically analyzed. In that subgroup, reader consensus had sensitivity of 83.3% and specificity of 90.9% for tumor progression compared with the pathologic result.
Relative SUV was a predictor of tumor progression and radiation effect after GKRS (univariate, p = 0.033; multivariate, p = 0.035). None of the other pretreatment demographic or clinical characteristics, including age, sex, lesion size and location, time between GKRS and PET, and radiation dose, was associated with response to therapy, individually or as a group (Tables 2 and 3).
Relative SUV had a diagnostic ROC AUC of 0.67 (95% CI, 0.55–0.79) (Fig. 1). Figure 2 shows the relation between sensitivity and specificity for identifying tumor progression diagnostic as a function of relative SUV. Relative SUV of 1.75 had sensitivity of 87.5% and specificity of 32%, and relative SUV of 2.75 had specificity of 92.5% and sensitivity of 22.5%. Figures 3 and 4 show representative contrast-enhanced MR, FDG PET, and FDG PET/MR images of two patients with brain metastases from primary lung cancer treated by GKRS.
 View larger version (44K) | Fig. 1 —Graph shows empiric and parametric ROC curves of diagnostic accuracy for predicting tumor progression as function of relative standardized uptake value. Empiric AUC is 0.67. |
 View larger version (45K) | Fig. 2 —Graph shows relation between sensitivity and specificity for identifying tumor progression as diagnostic function of relative standardized uptake value (SUV). Relative SUV of 1.75 has sensitivity of 87.5% and specificity of 32%; relative SUV of 2.25 has sensitivity of 54% and specificity of 72.5%; relative SUV of 2.75 has sensitivity of 22.5% and specificity of 92.5%. |
 View larger version (271K) | Fig. 3A —53-year-old woman with left parietal metastasis from primary lung cancer treated with gamma knife radiosurgery. A, Contrast-enhanced MR (A), FDG PET (B), and FDG PET/MR (C) images show left parietal enhancing lesion with relative standardized uptake value of 3. Biopsy revealed persistent poorly differentiated adenocarcinoma. |
 View larger version (289K) | Fig. 3B —53-year-old woman with left parietal metastasis from primary lung cancer treated with gamma knife radiosurgery. B, Contrast-enhanced MR (A), FDG PET (B), and FDG PET/MR (C) images show left parietal enhancing lesion with relative standardized uptake value of 3. Biopsy revealed persistent poorly differentiated adenocarcinoma. |
 View larger version (274K) | Fig. 3C —53-year-old woman with left parietal metastasis from primary lung cancer treated with gamma knife radiosurgery. C, Contrast-enhanced MR (A), FDG PET (B), and FDG PET/MR (C) images show left parietal enhancing lesion with relative standardized uptake value of 3. Biopsy revealed persistent poorly differentiated adenocarcinoma. |
 View larger version (290K) | Fig. 4A —59-year-old woman with left parietal metastasis from primary lung cancer treated with gamma knife radiosurgery (GKRS). A, Contrast-enhanced MR (A), FDG PET (B), and FDG PET/MR (C) images show left parietal enhancing lesion with relative standardized uptake value of 1.21. Size of lesion at follow-up MRI progressively decreased, consistent with response to GKRS. |
 View larger version (294K) | Fig. 4B —59-year-old woman with left parietal metastasis from primary lung cancer treated with gamma knife radiosurgery (GKRS). B, Contrast-enhanced MR (A), FDG PET (B), and FDG PET/MR (C) images show left parietal enhancing lesion with relative standardized uptake value of 1.21. Size of lesion at follow-up MRI progressively decreased, consistent with response to GKRS. |
 View larger version (251K) | Fig. 4C —59-year-old woman with left parietal metastasis from primary lung cancer treated with gamma knife radiosurgery (GKRS). C, Contrast-enhanced MR (A), FDG PET (B), and FDG PET/MR (C) images show left parietal enhancing lesion with relative standardized uptake value of 1.21. Size of lesion at follow-up MRI progressively decreased, consistent with response to GKRS. |
| Discussion | |
|---|
The degree of metabolic activity in the residual enhancing lesion at posttherapy FDG PET/MRI coregistration, as reflected by relative SUV, is associated with local tumor control after GKRS in patients with brain metastases. Our data suggest that this metabolic metric might have clinical prognostic value, because it may add value in the triage of patients with brain metastases that exhibit lesion growth at posttherapy MRI. Calculation of relative SUV in a treated enhancing lesion with PET/MRI coregistration may enable identification of patients with expected tumor progression who may benefit from additional GKRS or further therapeutic interventions.
Our data suggest that relative SUV less than 1.75 may be used to identify lesions likely to respond to GKRS despite manifesting radiation reaction. The data also show that relative SUV greater than 2.75 may be used to identify lesions that are probably going to progress despite treatment. Patients in the latter group may be considered for further early treatment, and those in the former group may still benefit from short-term follow-up observation [10]. Additional investigation to confirm the robustness of this approach is warranted. With the use of proposed thresholds when relative SUV is below or above certain cutoffs, FDG PET/MRI coregistration may provide additional information about the progression versus radiation effect dilemma, and this information may complement morphologic or physiologic information obtained from follow-up MRI. Very often, patients with primary cancers outside the CNS and metastatic brain disease undergo FDG PET. Thus, metabolic information on the brain parenchyma may be readily available for fusion with MRI and further quantification to assess for additional information regarding the response to SRS.
We used a quantitative approach to derive relative metrics from standard clinical FDG PET examinations. This method can be reproduced and used by other investigators and clinicians. The relative normalized values used in our study may be more prone to generalization than a strategy based on a single cutoff value, such as maximum SUV, which is more susceptible to patient, tracer dose, and hardware or software variability. Prior studies investigating the role of FDG PET in the distinction of tumor recurrence versus radiation effect were conducted with a visual all-or-none or semiquantitative approach and small and heterogeneous samples and did not benefit from MRI coregistration [21, 22]. Coregistration of FDG PET and MR images made identification of small enhancing foci found at FDG PET easier and evaluation of SUV in such areas more accurate.
Our study had limitations. First, we used a combination of histologic and follow-up MRI findings to define response to therapy. In a prior cohort of patients with primary brain tumors, we used overall survival instead of the criteria used in this study to define outcome. We thought survival would be biased by other factors, such as primary cancer type and burden of extracranial metastatic disease, because systemic disease is the ultimate cause of death of most patients with brain metastases. Nonetheless, more than 20% of our target lesions were biopsied, and among those, reader consensus showed significant agreement with pathologic result. Second, advanced MRI techniques, such as perfusion imaging and spectroscopy, were not consistently used in our patient sample, so the role of these techniques was not evaluated in this investigation. Third, subjectivity and reproducibility of manual selection of ROIs instead of automatic segmentation may be an additional potential limitation. Finally, we might have incurred selection bias, because some patients probably went straight to repeat GKRS after receiving a hypermetabolic PET result.
| Conclusion | |
|---|
Quantitative relative SUV at posttherapy FDG PET serves as a biomarker of local response to SRS in patients with brain metastases who have lesion enlargement identified at follow-up MRI. These prognostic data may affect management, supporting the need for further therapeutic actions for selected patients.
| References | |
|---|
