|Home | About | Journals | Submit | Contact Us | Français|
Therapy-induced calcifications in glioblastoma are rarely recognized. They may represent regressive changes in the tumor tissue, but their occurrence and possible predictive or prognostic value have not been systematically assessed. The observation of hyperintense lesions on precontrast T1-weighted magnetic resonance images (MRIs) in 2 index patients with glioblastoma after therapy with bevacizumab, subsequently identified as calcifications on computed tomographs (CTs), prompted us to prospectively screen for these radiographic changes.
Therefore, 36 patients with recurrent glioblastoma prospectively treated with bevacizumab in an observational trial were examined every 8 weeks by MRI and, if clinically necessary, by CT. In 22 patients (61.1%), T1 hyperintense lesions became apparent after bevacizumab treatment. The median time to detection of these lesions was 55 days. In 14 (63.6%) of 22 patients, CTs were available and confirmed the existence of tumor calcifications. No substantial changes in T1 hyperintense lesions or calcifications were recognized on additional MRI or CT scans. Interestingly, the patients with therapy-induced T1 hyperintense lesions had better durations of progression-free survival than patients without these changes (median, 5.8 vs 3.5 months; P< .001), and the duration of overall survival was also superior (median, 9.7 vs 5.0 months; P= .006). There was a striking correlation between the appearance of therapy-induced T1 hyperintense lesions and overall response to bevacizumab. Therefore, this phenomenon is a rather early and time-limited event during the first weeks of treatment and appears to be response related.
In summary, T1 hyperintense lesions are common in patients with glioblastoma who have been exposed to bevacizumab, may represent a novel biomarker of response and outcome, and seem to correspond to tumor calcifications.
Bevacizumab, a vascular endothelial growth factor (VEGF)–neutralizing antibody, is the prototype of an antiangiogenic agent and has broad antitumor activity. Recently, bevacizumab was approved by the US Food and Drug Administration for the use in recurrent glioblastoma on the basis of unprecedented rates of response and 6-month progression-free survival (PFS). However, the impact of bevacizumab on overall survival (OS) is still debatable, and no predictive biomarker has yet been firmly established.1–4
Several studies have addressed the significance of diverse radiographic changes seen in patients with glioblastoma treated with bevacizumab, to better understand its mode of action and to predefine patients who will benefit from bevacizumab treatment.5–13 Reduced perfusion, enhanced tumor hypoxia, and tumor starvation due to nutrient depletion would be expected from antiangiogenic therapy, but anti-VEGF and anti-VEGF receptor therapy have been shown to initially normalize the tumor vasculature in preclinical models, with reduced vascular permeability, decreased interstitial pressure, and increased tumor oxygenation as consequences.14 This phenomenon has been detected in patients with glioma treated with the pan-VEGF receptor tyrosine kinase inhibitor AZD2171.15 Pope et al.8 recently reported that apparent diffusion coefficient (ADC) histogram analysis can predict for PFS in patients with recurrent glioblastoma who are treated with bevacizumab. The authors hypothesized that bevacizumab treatment is more effective for necrotic versus nonnecrotic tumors because they found a greater increase in median time to progression in bevacizumab-treated patients with high basal ADC values, possibly reflecting diminished cellular integrity in necrotic tumor areas. It has also been shown that changes in ADC values as early as 3 weeks after the start of treatment can predict response at 3 months. ADC values increased in patients subsequently responding to bevacizumab treatment and decreased in nonresponders.12 Regarding tumor perfusion, a more recent study showed that local perfusion changes, measured by change in hyperperfusion volume (ΔHPV), correlate better with time to progression than commonly used radiographic parameters do.13 In an [18F] Fluorothymidine (FLT) PET–based study, Chen et al.6 were able to show that a reduction of metabolic activity in an early FLT-PET 1–2 weeks after the start of treatment with bevacizumab seems to be predictive of OS in patients with recurrent glioma.
Here, we report on the occurrence of hyperintense lesions on precontrast T1-weighted magnetic resonance images (MRIs) in patients with glioblastoma after therapy with bevacizumab that correspond in several cases to calcifications in CTs. These lesions are common in patients with glioblastoma who are exposed to bevacizumab and may represent a novel biomarker associated with response and improved outcome.
This observational study was approved by the institutional review board (ethics committee at the University Hospital Frankfurt; reference number 4/09-SIN 01/09). Enrollment was restricted to patients with the histological diagnosis of a glioblastoma or gliosarcoma, radiologically confirmed recurrence, adequate laboratory values, and the recommendation of bevacizumab-based therapy. Because this was a noninterventional study, the decision on the individual treatment schedule and on any concomitant therapy (eg, chemotherapy and radiotherapy) was solely the responsibility of the treating physician. All patients were treated in our outpatient unit and were seen every other week. Thirty-six patients were enrolled in this study during the period of June 2008 through June 2010.
MRI was performed at baseline and in 8-week intervals using routine imaging parameters on a 1.5 T system (Intera; Philips Medical Systems) or a 3 T system (Trio; Siemens). The core protocol was identical for all patients, including transversal T2-weighted images (TR/TE, 4000/99 ms) and fluid-attenuated inversion recovery (FLAIR; TR/TI/TE, 10,000/2000/120 ms). For detection of contrast enhancement, T1-weighted images (TR/TE, 650/15 ms) were acquired before and after intravenous application of gadolinium-diethylene triamine pentaacetic acid (Gd-DTPA, 0.1 mmol/kg). Response was determined according to the updated response assessment criteria for high-grade gliomas (RANO) as complete response (CR), partial response (PR), stable disease (SD), or progressive disease (PD), except the fact that responses were not routinely confirmed by MRI 4 weeks thereafter. For the statistical evaluation, a CR or PR at any timepoint was considered a response even if the subsequent MRI 8 weeks later showed PD.
CTs can no longer be accepted as a sufficient standard of care to evaluate tumor status for patients with brain tumor. Therefore, CTs were only irregularly available in case of emergency situations, for treatment planning before radiotherapy or after surgical interventions. In some cases, CTs were scheduled to exclude the existence of hemorrhage during bevacizumab treatment. Thirty-one of 36 patients underwent CT before treatment with bevacizumab and 21 underwent CT during or after bevacizumab. Overall, 19 patients (52.8%) underwent CT before and after commencement of treatment.
The precontrast T1-weighted images of each time point were jointly analyzed for T1 hypintensities by 1 experienced neuroradiologist (E.H.) and 1 experienced neuro-oncologist (O.B). The positive evaluation criteria for “T1 hyperintense lesion” were as follows: (1) a new T1 hyperintensity versus the baseline MRI performed before treatment with bevacizumab; and (2) an unambiguous increase of T1 hyperintensity, when compared with the baseline MRI, of ≥200%.
Analyses of PFS and OS after the start of bevacizumab treatment were performed with the Kaplan-Meier method using the log-rank test on SPSS Statistics, version 17.0. The correlation between overall response (CR + PR) and the appearance of hyperintense lesions in the precontrast T1-weighted images was calculated using the Fisher exact test.
Patient characteristics and relevant pretreatment modalities are shown in Table 1. Distribution of age, sex, and Karnofsky performance score were largely similar to those of the hitherto largest trial on bevacizumab in patients with glioblastoma.1 All patients had received the histological diagnosis of a World Health Organization (WHO) grade 4 glioma. Three patients (8.3%) had a glioblastoma with an oligodendroglial component, 2 (5.6%) had a secondary glioblastoma, and another 2 (5.6%) had a gliosarcoma. The figures on tumors with an oligodendroglial component match the ones of previous studies, ranging from 5% to 17%.16–19 In the largest study, Salvati et al.18 found 36 tumors with an oligodendroglial component in a cohort of 450 patients with glioblastoma, corresponding to a proportion of 8.0%. Primary therapy generally consisted of radiochemotherapy, in accordance with the European Organisation for Research and Treatment of Cancer 26981/National Cancer Instiute of Canada CE.3 protocol. The median number of recurrences and the median number of prior chemotherapy treatments were 2. At recurrence, 8 patients had at least 1 resection, and 6 patients received reirradiation. The median time from diagnosis of glioblastoma to the start of bevacizumab treatment was 11 months. No other antiangiogenic agents were administered before or together with bevacizumab.
All patients received bevacizumab at a dose of 10 mg/kg body weight intravenously every other week. Table 2 shows the concomitant therapy. Fifty percent of the patients received bevacizumab alone, whereas in 41.7%, irinotecan was concurrently administered at a dose of 125 mg/m² because no patient was treated with enzyme-inducing antiepileptic drugs. In 4 patients, bevacizumab treatment was combined with temozolomide, lomustine, or reirradiation, and 1 patient received bevacizumab, irinotecan, and reirradiation.
The overall response rate was 66.7%, although 2 patients where lost to follow-up. Median time to best response was 53 days (Table 3). Median PFS was 4.7 months (95% confidence interval [CI], 3.7–5.7), and the 6-month PFS was 30.4% (Fig. 1A). Median OS was 8.3 months (95% CI, 6.4–10.2), and the 12-month OS was 24.9% (Fig. 1B). OS strongly depended on the type of best response (Fig. 1C). At therapy failure, 38.9% of patients had a progression of the contrast-enhancing tumor and 36.1% of the T2 area. Only 2 patients (5.6%) had progression with nonmeasurable disease, and no patient had progression with a new lesion. To date, 5 patients have not yet experienced progression (Table 3).
Of all patients, 22 (61.1%) developed new or significantly increased T1 hyperintense lesions in the tumor area. The median time to detection of the T1 hyperintense lesions was 55 days (Table 3). In 14 (63.6%) of 22 patients, suspected calcifications were confirmed by CT (Hounsfield units, >500). Only 1 patient with a new T1 hyperintense lesion did not have calcification revealed by CT. Of the 12 patients without T1 hyperintense lesions, 4 had CTs without calcifications, but incipient calcifications were detectable in 2 others. The radiographic observations of 4 representative patients are shown in Fig. 2. These cases illustrate the appearance of new T1 hyperintense lesions (Fig. 2, column “T1”), the response of the contrast enhancing tumor (Fig. 2, column “T1/Gd-DTPA”), and considering the different angle and slice thickness of CT and MRI, the development of calcifications in the same tumor region on the corresponding CTs (Fig. 2, column “CT”). A more detailed analysis of the follow-up situation for patient 4 from Fig. 2 is shown in Fig. 3. The bottom panel shows an overlay of the noncontrast T1 images (upper panel) and the calcified areas (red) of the noncontrast CT images (middle panel). The radiographic findings for another 6 patients after exposure to bevacizumab are shown in the supplementary figure S1.
Because of the missing genuine clinical indication, CTs as a reference standard for the detection of calcifications were not performed on a regular basis. Therefore, CTs were only irregularly and incompletely available. More than 85% of all patients had a CT at any time before start of bevacizumab treatment. Although it was done whenever clinically necessary during or after bevacizumab treatment, CTs before and after treatment start were only available for 19 patients (52.8%). The median time to occurrence of calcifications on CTs was 163 days with a range of 43 to 393 days (Table 3). However, because CTs were not part of the protocol, this information does not reflect the onset of calcifications accurately. Therefore, we correlated outcome variables with the T1-hyperintense lesions.
Oligodendroglial differentiation (Table 1) was not a predictor of T1 hyperintense lesions or calcifications on CTs. The proposed mechanism of bevacizumab-induced calcifications—that is, mineralizing angiopathy—is probably different from that of tumor calcifications in oligodendroglial tumors.
On neither follow-up MRIs nor the few available additional CTs, substantial changes in T1 hyperintense lesions or calcifications were recognized. Thus, the phenomenon of bevacizumab-induced T1 hyperintense lesions/calcifications rather seems to be an early and time-limited event during the first weeks.
There were no gross differences in distribution of age, sex, Karnofsky performance score, pretreatment, and number of recurrences between patients with or patients without new or significantly increased T1 hyperintense lesions. Reirradiation before or with bevacizumab especially had no significant influence on the formation of T1 hyperintense lesions.
The median PFS times were 3.5 months (95% CI, 2.9–4.1) for patients without and 5.8 months (95% CI, 4.0–7.6) for patients with new or significantly increased T1 hyperintense lesions (P< .001). The 6-month PFS rates were 9.7% and 41.1%, respectively (Fig. 4A).
The median OS durations were 5.0 months (95% CI, 2.4–7.5) and 9.7 months (95% CI, 6.6–12.7) for patients without or with new or significantly increased T1 hyperintense lesions, respectively (P= .006). The respective12-month OS rates were 9.4% and 33.0% (Fig. 4B).
Because T1 hyperintense lesions seemed to preferentially appear in patients responding to bevacizumab treatment (CR + PR), we looked for a correlation between these 2 factors. A Fisher exact test revealed a significant correlation between the appearance of T1 hyperintense lesions and objective response (P < .0001) (Table 4).
Antiangiogenic therapy with bevacizumab results in high radiologic response rates and prolongs PFS in patients with glioblastoma.1–4,20 However, the mechanisms of bevacizumab action are not completely understood. Increasing interest in radiographic findings in bevacizumab-treated patients has developed in recent years with the particular aim of finding prognostic surrogate markers.
The present study identified T1 hyperintense lesions on the precontrast MRI as a marker of better outcome in patients with glioblastoma treated with bevacizumab. Of the 36 included patients, 22 (61.1%) showed new or markedly increased T1 hyperintense lesions after bevacizumab treatment. Microhemorrhages as a cause of these lesions were unlikely because the signal intensity of these lesions did not change on additional MRIs. Even if hemorrhages can persist for months in individual cases, the T1 hyperintense lesions of intracellular or extracellular methemoglobin should disappear with advancing hemoglobin degradation and hemosiderin accumulation.21 However, 14 (63.6%) of these 22 patients had CTs without evidence of hemorrhage and confirming the existence of tumor calcifications in the region of the T1 hyperintense lesions. Finally, we exclude the possibility that T1 shortening was associated with subacute hemorrhage, but the correlation with calcifications on CT and the persistence of the MRI signal increase over time strongly argue against this thesis. Regarding the colocalization of alterations on CT with those on MRI, Figs. Figs.2,2, ,3,3, and S1 illustrate that the T1 hyperintense lesions are confined to the region of calcifications and vice-versa, notwithstanding subtle differences in angulation between CT and MRI and the time intervals between acquisition of MRIs and CTs. These very probably reflect the same bevacizumab-induced process. Nevertheless, there is some heterogeneity within the lesions, with some areas more prominent on CT and others more prominent on MRI. We hypothesize that this reflects different stages of the calcifying process.
The T1 hyperintense lesions were apparent after a median time of only 55 days, (ie, on the first post-treatment MRI), and then remained stable. Therefore, the development of these lesions is not simply a consequence of prolonged treatment with bevacizumab, but rather an early and time-limited event during the first weeks of treatment. Because the appearance of T1 hyperintense lesions was highly correlated with objective response to bevacizumab, it appears plausible that these lesions are a result of bevacizumab-induced biological processes in the tumor tissue.
Usually, calcifications are easily depicted on normal CTs, but MRI has replaced CT in the follow-up of patients with primary brain tumors. In contrast, the detection of calcifications on standard MRI sequences is more difficult and the signal alterations are heterogeneous. Hyperintense lesions on nonenhanced T1-weighted MRIs in the brain (areas of T1 shortening) may be related to the presence of intra- or extracellular methemoglobin,22–24 as well as to the deposition of minerals, such as calcium and manganese,23,25–27 cholesterol,28,29 fat,30 melanin,31 or proteinaceous material28,32 or to free radical generation.33,34 High signal intensities from calcified areas may be explained by the shortening of the T1 relaxation time of protons next to the surface of the calcium crystals.35 It could be shown that the hyperintense T1 signal corresponds to calcifications depicted on cranial CT, even though the extent of signal increase may be smaller or even may be missed.25,35,36 Standard T2* weighted images are sensitive to the susceptibility effects of calcifications; however, the detection of T2* signal decrease is quite unspecific. Even the advantages of not-so-widespread susceptibility-weighted imaging (SWI) can finally not replace CTs as the gold standard. SWI is a T2* imaging process that includes phase information that makes it possible to differentiate paramagnetic substances (eg, deoxyhemoglobin, hemosiderin, and ferritin) from diamagnetic calcium. Because we used a common standard MRI protocol with “simple” T2* imaging lacking phase information, SWI was not part of the protocol. Congruent with our findings, high T1 signal intensities from calcified areas, depending on the calcium concentration and on the surface of the calcium particles, have been reported.25,27,35–37 Because T1 hyperintense lesions in precontrast MRI turned out to be the most feasible marker for the observed tumor calcifications in our cohort, they were used for additional analyses.
Cerebral calcifications are known to occur in some types of brain tumors. They are most common in oligodendrogliomas, but rare in astrocytic neoplasms, including glioblastoma. Importantly, calcifications in the aforementioned tumors occur spontaneously. In addition, the proportion of patients with tumors displaying a oligodendroglial component in our cohort is low (8.6%) and matches the ones from previous studies, thus excluding the possibility that this might affect the study results. Furthermore, although oligodendroglial tumors are associated with calcifications, it is noteworthy that this association is valid only with regard to the incidence of spontaneously occurring calcifications. We are not aware of work describing therapy-induced calcifications in the literature in oligodendroglial tumors. Thus far, therapy-induced calcifications have only been recognized as rather long-term sequelae of radiotherapy, typically for childhood tumors, such as medulloblastoma. The pathogenetic mechanisms of postirradiation cerebral calcifications are also unclear. However, it has been proposed that these are due to mineralising microangiopathy.38 On the basis of reports on the effects of irradiation on cerebral blood vessels,39 it has also been hypothesized that irradiation-induced vasculopathy results in hypoxia, with hypoxia-induced tissue damage then presenting as calcification.40 This is intriguing, given the potential of bevacizumab to induce both vessel regression and hypoxia in glioblastomas.15,41,42 Furthermore, cerebrovascular/ischemic events and the postanoxic state can be associated with cerebral calcifications.43,44 Another disease showing cerebral calcifications is the MELAS syndrome (mitochondrial myopathy, encephalopathy, lactic acidosis, and strokelike episodes).45 Sue et al.45 found that the common basal ganglia calcifications in that syndrome were found to be located in the pericapillary regions of the globus pallidus, with no neuronal involvement.
In addition to the phenomenon of vascular normalization, there have been inconsistent reports whether or when bevacizumab treatment results in augmented tumor hypoxia. Two recently published studies by de Groot et al.41 and our group suggest that bevacizumab treatment can lead to hypoxia.10,11 Bevacizumab possibly induces both early vascular normalization in the whole tumor volume and chronic hypoxia at least focally.
In the current study, we were not able to define the mechanisms underlying the formation of T1 hyperintensities. However, the available CTs and the characteristics and time course of the MRI findings are highly suggestive of a calcificating process. Regarding some of the supposed mechanisms in other forms of brain tissue calcifications (postanoxic calcifications, postirradiation calcifications, and calcifications associated with MELAS syndrome) with a central role for blood vessels and hypoxia, a link between blockade of VEGF and the formation of calcifications due to therapy-induced changes in tumor blood vessels appears possible.38–40
In conclusion, this is the first report on bevacizumab-induced T1 hyperintense lesions, or rather tumor calcifications, in patients with glioblastoma that are related to a better outcome. We hypothesize that there might be a link between bevacizumab-induced changes in tumor blood vessels and/or focal induction of hypoxia and the formation of these tumor calcifications. In addition, this study reveals that hyperintense lesions on precontrast T1-weighted MRIs in this patient population do not necessarily correspond to microhemorrhages, and tumor calcifications should be kept in mind as a cause of these imaging findings. New T1 hyperintense lesions or tumor calcifications detected on CTs in bevacizumab-treated patients may also prove to be valuable as a predictive biomarker in this patient population.
The Dr. Senckenberg Institute of Neurooncology is supported by the Dr. Senckenberg Foundation and the Hertie Foundation. J.P.S. is Hertie Professor of Neurooncology.
Conflict of interest statement. J.P.S. has served as a consultant for Roche, the European distributor of bevacizumab (Avastin). The other authors have no conflicts of interest.