Relevance of T2 signal changes in the assessment of progression of glioblastoma according to the Response Assessment in Neurooncology criteria

Abstract

Background: According to the Response Assessment in Neurooncology (RANO) criteria, significant nonenhancing signal increase in T2-weighted images qualifies for progression in high-grade glioma (T2-progress), even if there is no change in the contrast-enhancing tumor portion. The purpose of this retrospective study was to assess the frequency of isolated T2-progress and its predictive value on subsequent T1-progress, as determined by a T2 signal increase of 15% or 25%, respectively. The frequency of T2-progress was correlated with antiangiogenic therapy. Patients and Methods: MRI follow-up examinations (n = 777) of 144 patients with histologically proven glioblastoma were assessed for contrast-enhanced T1 and T2-weighted images. Examinations were classified as T1-progress, T2-progress with 15% or 25% T2-signal increase, stable disease, or partial or complete response. Results: Thirty-five examinations revealed exclusive T2-progress using the 15% criterion, and only 2 examinations qualified for the 25% criterion; 61.8% of the scans presenting T2-progress and 31.5% of the scans presenting stable disease revealed T1-progress in the next follow-up examination. The χ² test showed a highly significant correlation (P < .001) between T2-progress, with the 15% criterion and subsequent T1-progress. No correlation between antiangiogenic therapy and T2-progress was shown. Conclusion: Tumor progression, as determined by both contrast-enhanced T1 and T2 sequences is more frequently diagnosed than when considering only contrast-enhanced T1 sequences. Definition of T2-progress by a 15% T2-signal increase criterion is superior to a 25% criterion. The missing correlation of T2-progress and antiangiogenic therapy supports the hypothesis of T2-progress as part of the natural course of the tumor disease.

Glioblastoma is the most frequent malignant primary brain tumor, accounting for 12% to 15% of all intracranial neoplasms.^1–4 Despite recent therapeutic advances, the prognosis for patients with glioblastoma remains dismal with a median overall survival (OS) of 12–14 months.^5,6 Therefore, the need for more effective therapies is as evident as the need for controlled clinical studies and study-specific end points.⁷

Progression-free survival (PFS), radiographic response rate (RR), and OS are the most commonly used end points in clinical studies. In contrast to OS, both PFS and RR rely on accurate and reproducible methods to determine tumor progression.⁸ With more effective therapies to come, the definition of progression also influences the decision whether and when to switch from one treatment to another and may therefore guide relevant daily clinical decisions.

Traditionally, treatment response of brain tumors was assessed using the Macdonald criteria, which defined progressive disease (PD) by at least 25% increase in the sum of the products of perpendicular diameters of the enhancing lesion or appearance of any new lesion.^9,10 With a few exemptions, Macdonald criteria have been widely used in studies in glioblastoma since their introduction.¹¹

Macdonald criteria exclusively refer to the contrast-enhancing tumor portion. This assumption constitutes an important limitation, because the enhancing tumor areas primarily reflect passage of contrast agent across a disrupted blood brain barrier (BBB) and not necessarily viable tumor tissue.^8,12 It is widely accepted that there are numerous factors influencing the permeability of the BBB that are not associated with tumor activity.^7,13–17

Vice versa, high-grade gliomas are highly infiltrative in nature, and their diffuse tumor growth does not necessarily result in a disruption of the BBB.¹¹ Therefore, nonenhancing infiltrative progress, which is best detectable on T2 fluid–attenuated inverse recovery images (FLAIR;¹⁸ T2-progress) can occur without any change or even a decrease of the enhancement on T1-weighted images. Older reports estimated that T2-progress accounted for 10% of all recurrent cases of glioblastoma. T2-progress is supposed to occur more frequently in patients treated with antiangiogenic therapy,^19,20 although this has recently been under debate.^21,22 Macdonald criteria did not consider T2-progress.

Because of the limitations of Macdonald criteria, the Response Assessment in Neuro-Oncology (RANO) Working Group has recently published new recommendations for the assessment of treatment response in high-grade glioma (Table 1).¹¹ As in the Macdonald criteria, T1-progress is defined as an increase of at least 25% in the sum of perpendicular diameters of enhancing lesions. In contrast to the Macdonald criteria, the updated criteria include T2-progress, independent of contrast-enhanced T1-weighted sequences. T2-progress and, consequently, PD is assumed if nonenhancing lesions on T2/FLAIR-weighted images increase significantly on a stable or increasing dose of corticosteroids, compared with baseline scan or best response after initiation of therapy.

Table 1.

Summary of the RANO Response Criteria [14].

Criterion	CR	PR	SD	PD
T1 gadolinium enhancing disease	none	≥50% ↓	<50% ↓ but <25% ↑	≥25% ↑a
T2/FLAIR	stable or ↓	stable or ↓	stable or ↓	↑a
New lesions	none	None	None	presenta
Corticosteroids	none	stable or ↓	stable or ↓	NAb
Clinical status	stable or ↑	stable or ↑	stable or ↑	↓a
Requirement for response	all	All	All	anya

Abbreviations: RANO, Response Assessment in Neuro-Oncology; CR, complete response; PR, partial response; SD, stable disease; PD, progressive disease; FLAIR, Fluid-attenuated inversion recovery; NA, not applicable.

^aProgression occurs when this criterion is present.

^bIncrease in corticosteroids alone will not be taken into account in determining progression in absence of persistent clinical deterioration.

One of the major challenges associated with the T2 assessment according to the RANO criteria is the differentiation between T2-progress and other causes of T2 signal increase, such as edema, radiation effects, decreased corticosteroid dosing, seizures, postoperative changes, or even ischemic injury.^11,23 It is still unclear the number of cases in which these causes of T2-signal increase may imitate T2-progress.

Furthermore, an obvious limitation of the T2-progress assessment according to RANO criteria is the missing quantification of the required T2 signal increase. Generally, T2-progress is considered if changes in the T2/FLAIR signal suggest infiltrating tumor, including mass effect and infiltration.¹¹ The RANO working group justified this restriction with the limitations of the given technology.

The aim of this study was to evaluate in a large glioblastoma cohort the relevance of including T2 images in the assessment of tumor progress, compared with progress assessment based on tumor enhancement in T1-weighted images alone. Furthermore, we wanted to assess whether T2-progress predicts upcoming T1-progress. Therefore, we hypothesized that T2-progress would be more frequently followed by T1-progress in the subsequent follow-up examination than scans that revealed stable disease (SD). If T2-progress was not followed by T1-progress in the subsequent follow-up, it was analyzed whether the T2-signal increase remained or disappeared until T1-progress was diagnosed at a later follow-up examination (Fig. 1). To assess the practicability of RANO criteria in daily clinical decision-making, we operationally defined 2 thresholds of T2-progress, with an increase in the area of elevated T2 signal by 15% and 25%, respectively. Subsequently, the number of scans revealing a T2-signal increase of 15% or 25% were compared. Finally, we analyzed whether the patients receiving antiangiogenic medication on the scan date were more likely to be assessed as T2-progress.

Fig. 1.

In 21 (61.8%) of 34 scans, initial T2-progress was followed by T1-progress in the subsequent follow-up. In 12 (92.3%) of the remaining 13 cases, the T2-signal increase remained and T1-progress was diagnosed at a later follow-up visit. Only in one case, T2-signal increase decreased without T1-progress in subsequent follow-up.

Material and Methods

Included MRI Scans

Included MRI scans were identified by the key words “glioblastoma,” “GBM,” or “glioma” in the in-house radiological information system of the Heidelberg University Medical Center from 1 January 2009 through 1 January 2011. All available follow-up examinations of patients with glioblastoma during this period were included. Prior scans of the identified patients starting from 1 January 2005 were also considered. A mean interval between MRI scans was calculated. All patients enrolled in this retrospective study underwent serial MRI in our institution. All patients were treated at the Medical Centers in Heidelberg and had consented to MRI and therapy according to German regulations. All patients fulfilled the following criteria: (1) pathology-confirmed glioblastoma and (2) baseline (presurgical or postsurgical) MRI and at least 1 follow-up visit 6–24 weeks after baseline MRI scan, including T2 and contrast-enhanced T1-weighted images.

Imaging

MRI was performed on a 3 Tesla scanner (Trio and/or Verio; Siemens Erlangen) or a 1.5 Tesla scanner (Symphony; Siemens Erlangen). The standard MRI protocol included axial T1-weighted imaging (TR, 400 ms; TE, 15 ms; section thickness, 5 mm; FOV, 230 mm) or a T1-weighted 3D MPRage (TR, 1740 ms; TE, 3.45 ms; section thickness, 1.0 mm; FOV, 250 mm) without and with contrast media (Gadovist; Bayer), T2-weighted imaging (TR, 4890 ms; TE 85 ms; section thickness, 5 mm; FOV, 230 mm), FLAIR (TR, 8500 ms; TE, 85 ms; TI, 2400 ms; section thickness, 5 mm; FOV, 230 mm), and diffusion-weighted imaging (TR, 3000 ms; TE, 86 ms; b = 0 s/mm², 3 × b = 1000 s/mm², section thickness 5 mm; FOV, 230 mm).

Determination of Tumor Progression

All follow-up examinations were assessed independently by 2 neuroradiologists (A.R. and K.L.) who were blinded to all patient data. Discrepancies were resolved by consensus reading. Each follow-up was evaluated as SD, PR, CR, T1-progress, or T2-progress. According to the RANO criteria, T1-progress was assumed if there was a 25% increase in the sum of the products of perpendicular diameters of enhancing lesions, compared with the smallest tumor measurement either at baseline or best response. Furthermore, any new enhancing lesion with a nodular component measuring ≥10 mm in diameter was evaluated as T1-progress according to RANO criteria.

In contrast to RANO criteria, T2-progress was defined as an increase in the area with T2 abnormal signal by at least 15% or 25%, compared with the smallest tumor measurement either at baseline or best response. Each T2-progress was evaluated as T2-progress showing a 15% signal increase or T2-progress showing a 25% signal increase. Furthermore, patients showing T2-progress were carefully evaluated to differentiate this finding from possible other causes. We evaluated the corticosteroid dosing on the basis of the notes of the physician who examined the patient on the scan date. We also assessed the clinical status and the presence of seizures or ischemic injury¹⁶ on the basis of the physician note on the scan date. The dates of radiation of the last tumor resection were determined. Subsequently, we assessed the frequency of follow-up examinations revealing SD or T2-progress followed by T1-progress. If T2-progress was not followed by T1-progress in the subsequent follow-up visit, we assessed whether the T2-increase subsided or increased until T1-progress appeared. Finally, we assessed overall survival of the patients who presented with T2-progress and compared overall survival data that resulted if T2-progress were chosen as baseline with the overall survival data that resulted if the subsequent T1-progress were chosen as baseline. If T2-progress appeared >1 times in a patient, the earliest T2-progress was chosen as baseline for the proposed analyses. Median time from T2-progress to the next follow-up scan to assess T1 progression was determined.

Determination of Medication at the Time of Tumor Progression

After determining the MRI scans presenting a T1 or T2 progression, the therapy regime at the given date on the basis of the physician note of the scan date was evaluated. We differentiated therapy regimes into 2 different groups: conventional chemotherapy (e.g., temozolomide, amino-chlorethyl nitrosourea [ACNU], bischloro-ethyl nitrosurea [BCNU]) and bevacizumab (BEV), an antibody to vascular endothelial growth factor (VEGF), the only antiangiogenic regime that was used in our patient cohort.

Statistical Approach

The probability that T2-progress showed subsequent T1-progress was compared with the probability that SD showed subsequent T1-progress in the next follow-up, using the χ² test. Furthermore, the association between T2-progress/T1-progress and antiangiogenic treatment was analyzed using the χ² test.

Results

A total of 777 follow-up examinations of 144 patients who received a diagnosis of glioblastoma (84 male, 60 female) were assessed. The mean number of scans per patient was 6.7 (standard deviation, 3.32; range, 1–23; median, 5). The median age of the patients was 57 years (mean, 55 years; range 20–84 years).

The mean interval between scans was 84.45 ± 35 days. Progressive disease was diagnosed in 363 follow-up visits, of which 35 (9.6%) were evaluated as exclusive T2-progress, 15 (4.1%) as exclusive T1-progress, and 313 (86.2%) as combined T1- and T2-progress. Follow-up of 390 images revealed SD, 17 PR, and 7 CR (Table 2). One hundred thirty-eight of the included 144 patients presented with at least one T1-progress during the investigated period. The remaining 6 patients were reviewed until June 2011. Three of them presented with T1-progress during this period.

Table 2.

Assessment of 777 f/u examinations in 144 patients

SD	390
PR	17
CR	7
PD: Combined T1/T2 -progress	313
PD: exclusive T1-progress	15
PD: exclusiveT2- progress (15% + 25%)	33 + 2
Σ	777

Of the evaluated patients with T2-progress, 35 fulfilled the 15% criterion, and only 2 fulfilled the 25% criterion (Table 2). T2-progress was diagnosed in 30 patients. In one patient, T2-progress was diagnosed 3 times; in 3 patients, 2 times; and in 26 patients, T2-progress was diagnosed once during the course of the disease.

All follow-up examinations showing T2-progress were of patients receiving stable corticosteroids within 14 days prior to the scan date. Three of the identified 35 patients with T2-progress showed clinical deterioration and were started on steroids after the examination. None of the patients with T2-progress had clinically suspected ischemic injury or seizure within the 4 weeks prior to MRI, accounting for the increase in T2 lesions. Furthermore, none of the patients with T2-progress had surgery or radiation therapy during the 3 months prior to MRI. One MRI displaying T2-progress did not have a subsequent MRI for determination of subsequent T1-progress and, thus, was excluded from further analysis. Twenty-one (61.8%) of the remaining 34 patients with T2-progress presented additional T1-progress at the next follow-up examination, whereas 13 (38.2%) follow-up scans were stable on T1 images. Among these 13 patients, the T2-signal increase only subsided in 1 patient, whereas in the other 12 patients, the initial increase continued further until T1-progress appeared in the additional follow-up MRIs (Fig. 1). The median time from T2-progress to the next follow-up scan to assess T1 progression was 82 days.

Of the remaining 390 SD scans, 333 had subsequent follow-up, which revealed T1-progress in 105 scans (31.5%) (Figs 2–4). A χ² test showed a highly significant correlation (P < .001) between T2-progress and the probability of T1-progress in the subsequent follow-up.

Fig. 2.

Relevance of including T2 images according to RANO criteria in the assessment of tumor progress, compared with assessment according to Macdonald criteria. (A) One MRI of the 35 scans displaying T2-progress did not have a subsequent follow-up. Twenty-one of the remaining 34 T2-progresses (61.8%) presented additional T1-progress at the next follow-up. An example for T1-progress following T2-progress is given in Fig. 3. (B) Of the 390 SD scans, 333 had subsequent follow-up, which revealed T1-progress in 105 scans (31.5%). An example for T1-progress without previous T2-progress is given in Fig. 4. A χ² test showed a highly significant correlation (P < .001) between T2-progress and the probability of T1-progress in the subsequent follow-up; T2 progress predicts upcoming T1-progress.

Fig. 3.

T2-progress predicts T1-progress. (a and b), Recurrent glioblastoma with a contrast-enhancing tumor (b) and hyperintense changes surrounding the tumor on FLAIR images (a). (c and d), Follow-up (after 3 months) shows a progression of the hyperintense changes surrounding the tumor (c) with stable contrast-enhancing tumor (d). (e and f), Another 3 months later, the follow-up shows a progression of the contrast-enhancing tumor (f) and an increase of the hyperintense FLAIR changes (e).

Fig. 4.

T1-progress without previous T2-progress. (a and b), Recurrent glioblastoma after initial therapy with a contrast-enhancing tumor (b) and additional hyperintense changes on FLAIR images (a). (c and d), At 2.5 months follow-up, stable disease (stable contrast-enhancing tumor [d] and stable hyperintense FLAIR changes [c]). (e and f), The next follow-up 3 months later shows a progression of the contrast-enhancing tumor part and an increase of the hyperintense changes on FLAIR images, combined T1- and T2-progress.

Patients were treated with BEV in 4 of the assessed 35 T2-progresses, whereas patients were treated with BEV in 55 of 328 follow-up examinations presenting T1-progress. A χ² test did not reveal any correlation between BEV treatment and T2-progress (Table 3).

Table 3.

No correlation is shown between medication with BEV and T2-progress

	T1-progress	T2-progress
Bevacizumab	55	4	59
No Bevacizumab Conventional Chemotherapy	273	31	304
	328	35	363

Twelve of the 30 patients presenting with T2-progress were still alive when the database was closed on July 2011. Median OS of the deceased 18 patients who presented with T2-progress was 363 days (mean ± standard deviation, 373 ± 196 days) when the scan of T2-progress was chosen as baseline. OS of the patients who were still alive was censored when the database was closed on July 2011 (median, 412 days; mean ± standard deviation, 539 ± 247 days). Three patients in the first group and 1 patient in the group that was still alive at closure of the database had secondary glioblastoma. Scans of those patients were only included when the latest diagnosis was glioblastoma.

If the subsequent T1-progress would have been chosen as baseline for the determination of OS, the median would have been 248 days (mean ± standard deviation, 270 ± 92 days). PD was diagnosed a mean of 125 days earlier than when only T1-progress was considered.

Discussion

Our data confirm the hypothesis that the inclusion of T2-weighted sequences in response criteria of glioblastoma as suggested by the RANO criteria¹¹ results in relevant changes in response assessment. In this cohort  of patients with glioblastoma, 9.6% of MRI scans that were evaluated as PD according to the RANO criteria showed exclusive T2-progress. These examinations would have been rated as SD or PR without consideration of T2-progress. Clinical relevance of these findings is underlined by the fact that PD is diagnosed a mean of 125 days earlier than when only T1-progress is considered. Therefore, OS among patients is significantly longer if T2-progress is chosen as a start point and not the following T1-progress. On the other hand, 4.1% of MRI scans revealed T1-progress without T2-progress. Therefore, only the combined use of T1- and T2-progresses seems to be adequate in the tumor response evaluation.

However, the relevance of T2-progress for clinical decisions and for the course of the disease is unclear. Furthermore, there are open questions regarding the biology of T1- and T2-progress. It still remains unknown whether T2-progress and T1-progress represent 2 different phenotypes or whether invasiveness and corresponding T2-signal increase is part of the natural course of the disease predicting upcoming T1-progress.²¹

As mentioned above, T1-progress primarily reflects a disrupted BBB, whereas T2-progress is supposed to refer to the infiltrative growth pattern of glioblastoma. On a molecular basis, a possible explanation for T2-progress preceding subsequent T1-progress could be that migrating tumor cells in the invasion front, causing T2-signal increase, produce and secrete neoangiogenic factors, leading to a trailing behind of the pathologically vascularized, leaky tumor-border, representing T1-progress.

The fact that T2-progress is not routinely detected may be attributable to the rapid BBB disruption following tumor invasion. According to this hypothesis, follow-up presenting combined T1- and T2-progress may have presented only T2-progress at a distinct time, before additional T1-progress subsequently occurred. Therefore, T2-progress could be characterized as an intermediate step in the natural course of the tumor-disease between SD and T1-progress. Specific subgroups of glioblastoma, defined for example, by genetic alterations or treatment, may exhibit different periods between T2-progress and T1-progress.

Cases revealing T1-progress without preceding T2-progress may represent an aggressive subgroup with almost concomitant neoplastic infiltration and BBB disruption. On the other hand, cases that presented continuous increase of T2-signal changes in follow-up examinations and, finally, in T1-progress may form a subgroup with a highly invasive pattern.

The fact that T2-progress was only in 3 of 35 cases accompanied by clinical deterioration might further support the hypothesis of T2-progress representing an early sign of progressive disease, which manifests clinically as T1-progress at a later stage. Furthermore, it is in accordance with this hypothesis that OS among patients after diagnosis of T2-progress varies heavily (80–873 days) and that T2-progress may be diagnosed several times (in 1 patient, 3 times) in the course of the disease. If T2-progress describes an intermediate step in the natural cause of the disease, this step should be detectable both in earlier and in later stages of the disease.

It is still unclear whether antiangiogenic therapy, such as BEV, in a subgroup of patients induces or predisposes for T2-progress. Under antiangiogenic therapies, an increase of T2-progress up to 20% to 30% has been reported.^18,24 Apart from these clinical series and some preclinical studies,^25,26 there is also a strong biological rationale that has caused the perception of antiangiogenic agents promoting T2-progress. Because glioblastomas produce a large amount of VEGF and are highly vascular, blocking of VEGF or VEGF receptors should normalize tumor vasculature, which should result in a decrease in contrast enhancement. Furthermore, continous blockade of VEGF-induced vascular proliferation may, at least in some patients and in several preclinical models, promote tumor escape (e.g., by vascular co-option), which should manifest as radiographically infiltrative T2-progress.²⁰

Our investigations did not support the hypothesis of antiangiogenic therapy promoting T2-progress. However, this theory has come under debate recently. In a matched-pair analysis of patients with malignant glioma treated with BEV and patients treated without anti-VEGF-regimes, we did not find a specific propensity of BEV to induce T2-progress at recurrence of high-grade glioma.²²

However, our investigation underlines the necessity to consider T2-progress in trials of antiangiogenic agents and in trials of classical agents. Further prospective studies are necessary to investigate our hypothesis of T2-progress being an intermediate step in the natural course of the tumor-disease between SD and T1-progress.

The quantity of increasing T2-signal to be taken as a cutoff to determine T2-progress remains crucial for clinical practice and unsolved in the RANO criteria. A low value may decrease the specificity by increasing the risk of misdiagnosis of T2-progression because of possible other causes of subtle T2-signal increase, such as decreased corticosteroid dosing, radiation therapy, ischemic injury, seizures, or postoperative changes. On the other hand, a high value decreases the sensitivity by increasing the risk of not recognizing the T2 progression. Because the T2-signal increase of 15% that was operationally determined in our study for T2-progress only regressed in one case, whereas all other cases involved T1-progress, this cutoff may represent a reasonable threshold. A cutoff value of 25% did not seem to be appropriate, because only 2 T2-progresses presented a T2-signal increase of 25%. Therefore, the remaining 33 T2-progresses that were additionally detected when using a 15% cutoff value would not have been diagnosed. Future studies should evaluate which amount of T2-signal increase predicts progressive disease with maximal sensitivity and specificity. Volumetric analysis may further help to determine T2-progress more adequately.²⁴

Furthermore, novel imaging techniques, such as diffusion-, perfusion-, or susceptibility-weighted MRI should be evaluated for confirmation of T2-progress. However, none of these techniques has proven to substantiate T2-progress reliably, although diffusion imaging and apparent diffusion coefficient maps may be useful.²⁷ Therefore, a careful analysis of possible other causes underlying the T2-increase and additional confirmatory scans represent the only available method to determine T2-progress.

The results of this large retrospective data set should be verified in future prospective trials. In these trials, the application of novel MR techniques should be evaluated intensively. One first example may be an upcoming trial of the European Organization for Research and Treatment of Cancer (EORTC 26101), which includes a uniform imaging protocol, including central reading, and aims at validating the RANO criteria. Ultimately, with the existence of effective recurrence therapies, the correlation of exclusive T2-progress with clinical outcome will be necessary to further substantiate the use of this definition.

Conclusion

Assessing T2-progresses, introduced by RANO criteria increases the sensitivity in detection of glioblastoma progression. T2-progress predicts subsequent T1-progress in a substantial number of patients, indicating T2-progress as an intermediate step in the natural course of the disease between SD and T1-progress. Definition of T2-progress by a 15% T2-signal increase criterion is superior to a 25% criterion. No correlation between T2-progress and treatment with BEV was detected.

Conflict of interest statement. None declared.