Optimization of diagnostic performance for differentiation of recurrence from radiation necrosis in patients with metastatic brain tumors using tumor volume-corrected ¹¹C-methionine uptake

¹¹C-methionine (¹¹C-MET) positron emission tomography/computed tomography (PET/CT) has been used to diagnose primary and metastatic brain tumors [1‐3]. Tumor to normal tissue ratio (T/N ratio) is a useful metabolic parameter to differentiate tumor recurrence from radiation necrosis after gamma knife radiosurgery (GKR) [2], which is an important treatment modality for metastatic brain tumors [4].

The T/N ratio is a relative index that can be affected by variable factors. Previous studies have shown that the cutoff value of T/N ratio for differentiation of tumor recurrence from radiation necrosis depends on the histologic types, tumor sizes, and other parameters [2, 5, 6]. T/N ratio could especially be affected by tumor size because ¹¹C-MET uptake in small-sized tumors is underestimated through a partial volume effect (PVE) [5, 7]. Although a different cutoff value of the T/N ratio should be applied for diagnosis according to tumor size, there have been no studies regarding reference parameters for measuring tumor sizes to stratify the cutoff values of T/N ratio.

Metabolic tumor volume (MTV) usually represents the volume of metabolically active tumor in a reproducible manner [8, 9], which could be a potential reference parameter to measure tumor size. In this study, we investigated whether cutoff values of the T/N ratio can be stratified according to MTV and the resulting improvement in the diagnostic performance of ¹¹C-MET PET/CT for diagnosis of tumor recurrence in patients with metastatic brain tumors.

Among 77 lesions on ¹¹C-MET PET/CT, 51 showed recurrence, which was diagnosed after pathologic confirmation in nine lesions and radiological tumor control after repeated GKR in 42 lesions. Twenty-six lesions were radiologically defined as radiation necrosis showing stable or reduced volume without treatment.

Early detection of recurrence after GKR was important for deciding the treatment strategy in patients with metastatic brain tumor. However, it is difficult to differentiate between radiation-induced necrotic inflammation and active tumor recurrence after treatment because necrotic lesions show similar enhancement on gadolinium-enhanced MRI [4, 11, 12]. ¹¹C-MET PET/CT has played an important role in the diagnosis of tumor recurrence [1, 13], but it has limited application in evaluation of small lesions because of PVE, which results in underestimation of apparent ¹¹C-MET activity [5, 7, 14]. These limitations of ¹¹C-MET PET/CT could be critical in patients who underwent GKR, because it has often been used to treat small metastatic brain tumors (mostly less than 4 cm in maximal diameter) to avoid side effects such as radiation-induced inflammatory changes [4, 15].

The accuracy of ¹¹C-MET PET/CT in distinguishing between recurrence and radiation necrosis has been variously reported as showing 60–100% sensitivity and 75–100% specificity in brain tumors [16]. Additionally, optimal cutoff value of T/N ratio, which was defined as maximal ¹¹C-MET uptake in the lesion divided by that of the contralateral normal gray matter, has also been reported with variable range (1.4–1.9) in previous studies [2, 3, 6]. Our study showed that mean value of T/N ratio tended to decrease in both recurrence and radiation necrosis as tumor size decreased (Table 1). This suggested that different cutoff values of T/N ratio should be applied according to tumor size. However, the measured diameter on ¹¹C-MET PET/CT might have poor inter-observer agreement. Tumor diameter on MRI does not represent the exact tumor burden because it includes both tumor recurrence and radiation necrosis. For that reason, it was very important to find a reproducible parameter to measure tumor size.

In several studies, MTV on ¹¹C-MET PET/CT has been reported to represent the real tumor burden, which has been correlated with progression-free or overall survival in patients with brain tumor [13, 17]. Moreover, MTV has potential as a size criterion measured with reproducible manner, which applies a threshold of 1.3 times the mean ¹¹C-MET uptake of contralateral normal gray matter. In our study, MTV was used as a size criterion for tumor grouping, derived from the diameter (1 and 2 cm) of the lesions that could be affected by PVE [5]. As a result, cutoff values of T/N ratio for diagnosis could be optimized in each group classified by MTV. In particular, in the group with MTV ≤ 0.5 cm³, there was no recurrent lesion with T/N ratio more than the non-corrected cutoff value (1.61). Conversely, in the group with MTV > 4.0 cm³, there was no radiation necrotic lesion with T/N ratio more than the MTV-corrected cutoff value (1.85) and no recurrent lesion with T/N ratio less than 1.61. This explains why McNemar’s test could not report the differences in diagnostic accuracies between the MTV-corrected and non-corrected cutoff values in each group.

Some necrotic lesions have also been reported to have a high ¹¹C-MET uptake probably due to blood-brain barrier (BBB) disruption, inflammation, or reactive gliosis [18, 19]. Our study showed that T/N ratio of radiation necrosis also increased as MTV increased. Moreover, mean T/N ratio of radiation necrosis (1.65) was more than the non-corrected cutoff value (1.61) in the group with MTV > 4.0 cm³ (Table 2). Large necrotic tissues could have more false-positive uptakes of ¹¹C-MET through variable factors mentioned above than small necrotic tissues, which could reduce the specificity of ¹¹C-MET PET/CT. This could be another reason why application of different cutoff values of the T/N ratio according to tumor size is desirable for differential diagnosis between recurrence and radiation necrosis in metastatic brain tumor.

The T/N ratio could be also affected by factors other than tumor size. Pathophysiology of the necrotic lesion after therapy could be one of the possible reasons for the difference in cutoff values from previous studies, which showed that reactive gliosis was observed in necrotic tissues of gliomas but not of metastatic brain tumor [20, 21]. Terakawa et al. [2] showed that optimal cutoff values were different according to histologic types (1.41 for metastatic brain tumor vs. 1.58 for gliomas). Moreover, they also reported that ¹¹C-MET uptake of normal gray matter in patients who underwent conventional radiotherapy tended to be lower than that in patients who underwent stereotactic radiosurgery. We additionally tried to evaluate whether T/N ratio was different according to histology of the primary lesion. When we compared the T/N ratios of recurrent lesions in metastatic lesions from the lungs, T/N ratio showed a variable range according to lung histology (adenocarcinoma: 2.10 ± 0.98; squamous cell carcinoma: 2.78 ± 1.18; small cell carcinoma: 1.74 ± 0.29), although the difference was not significant. Further studies are necessary to elucidate the relationship among T/N ratio, tumor histology, and tumor size.

There is no consensus regarding which parameter between SUVmax and mean standardized uptake value (SUVmean) of the normal gray matter on ¹¹C-MET PET/CT is more useful to determine T/N ratio for differentiating between recurrence and radiation necrosis after radiation therapy in patients with brain tumors. Our study showed that the area under the ROC curve was 0.756 for T/N ratio used with SUVmean of the normal gray matter, which was lower than that used with SUVmax (0.790). In addition, the areas under the ROC curve of T/N ratio used with SUVmax were higher in each group classified by MTV. We adopted the following definition of T/N ratio: SUVmax of the lesion divided by that of normal gray matter. This may explain why optimal cutoff value of T/N ratio could be 1.23 in small lesions (MTV ≤ 0.5), although MTV was measured by applying a threshold 1.3 times the SUVmean of the normal gray matter from previous studies.

Compared to previous studies, our results showed relatively low negative predictive value (NPV). There could be several explanations. NPV was especially low in subgroups with small (MTV ≤ 0.5 cm³)- and large (MTV > 4.0 cm³)-sized lesions. Mean values of T/N ratio in both recurrence and radiation necrosis tended to be lower as tumor size decreased (Table 1). Low NPV could result from high prevalence of false-negative lesions although we had optimized cutoff value of T/N ratio in small sized tumors. Conversely, optimal cutoff value of T/N ratio in large-sized lesions was higher than that in all lesions from ROC curve analysis (1.85 vs. 1.61). High cutoff value of T/N ratio could also increase the false-negative results in large-sized lesions. Apart from diagnostic accuracy, several patients might be overdiagnosed to have recurrent lesions because 42 among 51 recurrent lesions were diagnosed with radiological tumor control after repeated GKR according to definition.

There were several limitations in this study. First, the retrospective nature of the study introduces the possibility of selection bias. Repeated GKR was done in patients with suspected recurrence, who seemed to lead a bias at the time of conducting ¹¹C-MET PET/CT. The proportion between recurrence and radiation necrosis could affect optimal cutoff value of T/N ratio especially in small- or large-sized lesions. Second, the final diagnosis of radiation necrosis or tumor recurrence was performed mainly with radiological follow-up. The treatment effect of chemotherapy to control the primary tumor was not excluded. Third, a phantom experiment was not performed to evaluate PVE to affect T/N ratio in this study. Instead, we tried to set acceptable size criteria to measure MTV from previous study [5].

Our study showed that optimal cutoff values of the T/N ratio should be different to maintain diagnostic accuracy regardless of tumor size. MTV has potential to be used as a reproducible parameter to reflect actual tumor size. Finally, MTV-corrected cutoff values of the T/N ratio can improve diagnostic performance of ¹¹C-MET PET/CT without decreasing specificity especially in small lesions. We expect our results to contribute to reproducible and standardized interpretation of ¹¹C-MET PET/CT.