Are radiation-induced cavernomas clinically relevant findings? Results from long-term follow-up with brain magnetic resonance imaging of childhood cancer survivors

In recent decades, the percentage of children who can be cured from cancer has increased steadily, resulting in a growing number of childhood cancer survivors worldwide.1 However, many of these childhood cancer survivors develop chronic health conditions emerging years to decades later as a consequence of cancer therapy.2 Compared with chemo- or immunotherapy, late effects of radiation often occur in the former treated field, leading to subsequent neoplasms.3 The cumulative radiation dose correlates with rates of subsequent tumors in the central nervous system (CNS) that include mostly meningiomas or gliomas.4

Furthermore, cranial radiotherapy is known to increase the risk of de novo formation of cerebral cavernomas (syn. cavernous angiomas, cavernous hemangiomas, cavernous malformation) as well, which also occur sporadically or following a familial etiology and represent benign venous malformations with low flow and low pressure.5, 6 However, while other radiation-induced neoplasms in most cases lead to rather clear-cut monitoring or therapy recommendations according to an interdisciplinary consensus, the handling approach for radiation-induced cavernomas (RIC) has yet to be defined, particularly in the context of the follow-up of childhood cancer survivors focused on here. Although the majority of cavernomas follow a complication-free course and do not require intervention, we are already aware of a tendency to symptomatic intracranial hemorrhage with potentially considerable clinical impact and even subsequent need for surgery.7, 8, 9, 10 But so far systematic investigations concerning the hemorrhage risk of RIC, in particular, are not available.

Brain magnetic resonance imaging (MRI) is the modality of choice for the detection and follow-up imaging of cavernomas.11, 12 In 1994, Zabramski et al. defined a grading system for familial cavernomas depending on MRI findings (Table 1).13 Nikoubashman et al. took up this classification in the recent past and investigated criteria to calculate prospective annual hemorrhage rates of cavernomas.14 The study showed high bleeding risks for Zabramski type I (29.8%), type II (20.1%) and a proposed new type V (23.1%) and low bleeding risks for type III and IV (3.4% and 1.3% respectively). But in their study patients with a history of irradiation were explicitly excluded.14

Thus it remains unclear whether Nikoubashman’s interpretation of the MRI appearance of cavernomas based on the Zabramski classification might be helpful for estimation of the hemorrhage risk of RIC. We would like to help clarify this and therefore set our sights on assessing for the first time the risk of hemorrhage of RIC in the long-term follow-up of childhood cancer survivors with brain MRI in this case series.

This study was approved by the Institutional Review Board of the University of Luebeck (14– 180, 18–087).

Patients

We retrospectively analyzed 36 childhood cancer survivors out of 252 patients who were examined in the late effects clinic at the University Medical Center Schleswig Holstein in Luebeck over a period of 6 years (03/2014–02/2020). We limited the analysis to patients with acute leukemia or brain tumors who had been treated with cranial radiotherapy (alone or as part of total body irradiation) and had received cranial imaging as part of their follow-up. Additionally, at least 2 brain MRI examinations (1 at the time of the initial tumor diagnosis and at least 1 in the follow-up) were required for inclusion in the study. The patients at the late effects clinic (i) were younger than 18 at cancer diagnosis, (ii) were at least 18 at first follow-up in the clinic and iii) survived at least 5 years after the end of cancer therapy. Exclusion criteria for the study were (i) active malignancy and (ii) lack of consent. See the flowchart of the patient selection process in Figure 1.

Figure 1

The aim of this study was to describe for the first time the course of RIC during long-term follow-up of childhood cancer survivors, those formerly treated with cranial radiotherapy in connection with the initial diagnosis of leukemia or brain tumor. We thus assessed the hemorrhage risk of RIC following Nikoubashman’s interpretation of the MRI appearance of cavernomas based on the Zabramski classification and examined the question of how to deal with RIC that frequently appear in MRI follow-up examinations of childhood cancer survivors. In our experience to date this seems to be distinctly more difficult than providing recommendations regarding extraaxial or intraaxial neoplasms (especially meningiomas or gliomas), which have already been the subject of extensive discussion in relevant literature as long-term consequences.4 Several guidelines, often based on a risk stratification system mainly taking into account the initial cancer diagnosis and the treatment received, recommend specific follow-up examinations to facilitate early diagnosis and treatment of these sequelae.15

Besides neoplasms, neurocognitive deficits in connection with pathological findings in brain MRI, such as leukoencephalopathy or brain atrophy, are reported in connection with childhood cancer survivors.16 Also our results showed 4 radiation-induced meningiomas, but no gliomas. There were variable findings of leukoencephalopathy, brain atrophy and several other comorbidities, though this was not a focal point of our studies.

Our study focuses on RIC. An assumed correlation between cranial radiotherapy and cavernomas in children was primarily described in 1994.17 In the following years this phenomenon was investigated extensively in several studies and summarized in review articles with a visible predominance in young patients treated for cancer.18, 19, 20, 21, 22, 23, 24 Furthermore, researchers also looked into the question of whether RIC differ from other, sporadically or familially occurring cavernomas regarding their natural course. Baumgartner et al. showed findings after irradiation radiologically and pathologically similar to sporadically occurring cavernomas.25 Cha et al., on the other hand, found differences in histological findings in a comparison between 10 patients with “de novo” cavernomas and 7 patients with a total of 8 RIC and regarded the findings after radiation more as inactive organizing hematomas in blood-filled spaces after tissue destruction and thus different from proliferation of malformed vasculature in sporadic cavernomas.26 According to a study with a specific histopathological focus based on 13 patients with RIC, there is, both histologically and with regard to immunohistochemistry, an overlap to non-radiation-induced cavernomas (5/13) as well as to organizing coagulum-like lesions (8/13), as described by Cha et al., so that one must assume two different etiological mechanisms in the pathogenesis.27

Concerning RIC, our results include a significant relation to the initial tumor entity and the cumulative radiation dose to the brain, as reflected by the occurrence of RIC without exception in patients treated for medulloblastoma with high cumulative radiation doses. In 2007 Lew et al. looked at precisely such patients who had received radiation therapy for medulloblastomas in childhood and described cavernomas in this case as common with an increasing incidence over time.28 They concluded that most of these lesions follow a benign course and do not require intervention.28 Furthermore, results that show a correlation between RIC and the cumulative radiation dose can be found in a more recent study focusing on adult cancer patients.29

With regard to the correlation between RIC and cranial radiotherapy, data on the time lag of the occurrence after radiation are additionally available. Following examination of 5 patients with cavernomas after cranial or craniospinal irradiation, Jain et al. described a latency interval between treatment and time of detection of cavernomas with a mean of 19.5 years.5 The latency that we found between cranial radiotherapy and detection of RIC with a median time period of 15.5 years matches these investigations (IQR 5–27). Moreover, Strenger et al. has published results that also focus on risk factors for the occurrence of RIC.30 The investigations encompassed 171 patients and showed that particularly children < 10 years of age at the time of irradiation have a higher risk for RIC, whose occurrence was documented here as taking place in a period of 2.9–18.4 years after radiation.30 Singla et al. examined the long-term course of irradiated patients and described for all of them an occurrence of cavernomas with a latency of 2–8 years – though only 5 patients with ALL were analyzed.31

Due to possible hemorrhage complications of RIC, the two latter studies very explicitly advocated imaging monitoring and Singla et al. described neurological symptoms and individual craniotomies because of RIC in 4 of 5 patients.30, 31 We did not document any symptomatic hemorrhages in connection with RIC in our patient group.

A number of studies have examined the hemorrhage risk of cavernomas in general. Flemming et al. calculated a prospective risk in this context in a group of 292 patients with cavernomas and consecutive hemorrhage.7 It was determined that prior hemorrhage is one of the factors that may predict future hemorrhage.7 One publication showed in a comparison of the features of RIC (32 patients) with other cavernomas (272 patients) no differences in symptomatic hemorrhages.32 However, here RIC proved to be more likely to occur multilocular.32 This matches our results, which predominantly showed multilocular and small lesions (Zabramski type IV). In particular, Nikoubashman et al. investigated those small, dot-like cavernomas in another study from 2013.33 It shows a low bleeding rate of 0.7% per lesion-year according to a survey specifically of 18 children.33 Nonetheless, it was propagated that even those small cavernomas are not purely benign lesions.33

It is certainly more difficult, however, to deal with lesions classified as Zabramski type I or II. As a reminder: type I is a (sub-)acute hemorrhage while type II as a “popcorn lesion” shows multiple hemorrhages. Here we again refer to data on the generally increased bleeding risk of cavernomas with already existing hemorrhage.7 Among others, Jeon et al. based their study of 326 adults with cavernomas, like us, on the Zabramski classification.34 The study showed that not only prior symptomatic hemorrhage, but also MRI appearance could be related to prospective symptomatic hemorrhages of cavernomas.34 The bleeding risk of cavernomas was also examined in a study by Al-Shahi Salman et al., which additionally evaluated gender-specific differences.35 This study found that the risk of recurrent hemorrhage is greater than the risk of a first bleeding event and greater for women than for men.35

As regards our results, we classified RIC as high risk for hemorrhage (Zabramski type I and II) in only 3 of 36 childhood cancer survivors (8%) and all other RIC were classified as low-risk lesions (Zabramski type IV [microhemorrhagic]). In this risk assessment we orient ourselves to the proposal made by Nikoubashman et al.14 After analysis of 199 MRI examinations in total with 1558 distinct lesions this team depicted 5 different types of cavernomas via 1.5T MRI (T1, T2 and T2*) while using a three-tier classification shown as graphic illustrations and corresponding MRI findings. It turned out that the cavernomas described as Zabramski type I, II and an additionally proposed new type V (with visible parts in the center of the actual cavernoma) had a significantly higher annual hemorrhage rate than type III and IV lesions.14 Thus, the presence of acute or subacute blood degradation was statistically the strongest indicator for an increased hemorrhage risk with the conclusion that it is possible to predict hemorrhage rates based on the Zabramski classification.14 As mentioned previously, these results were not related to RIC. Nevertheless, just in connection with RIC we consider the initial application of Nikoubashman’s interpretation of the MRI appearance of cavernomas based on the Zabramski classification for an assessment of the hemorrhage risk to be a very meaningful approach.

It must be emphasized that in this retrospective case series half of the childhood cancer survivors examined in long-term follow-up showed RIC in follow-up brain MRI. Thus, our study supports the development of cavernomas after cranial radiotherapy as a common side effect in childhood cancer survivors, though most RIC are not likely to hemorrhage and are detected incidentally. Our results therefore suggest a mainly benign course of RIC. Consequently we conclude that the complication risk of RIC can definitely be described in good conscience as low in communication with childhood cancer survivors as long as one applies assessments of existing classifications.13, 14 This ensures an advantageous aspect in dealing with patients who might feel anxious as a result of their prior cancer history. Studies with larger groups of patients and stronger study designs may possibly clarify whether dealing with RIC should be included and defined in the guidelines in connection with childhood cancer survivors.

Thus, we conclude that RIC are common late effects in childhood cancer survivors treated with cranial radiotherapy affecting half of the patients. However, only a few RIC (occurring in 8% of all reviewed childhood cancer survivors) were classified as high risk for hemorrhage and the majority of RIC follow a benign course. None of the childhood cancer survivors with RIC developed symptomatic hemorrhages. These results augment accurate risk assessment in individualized long-term follow-up also in order to reduce morbidity in childhood cancer survivors. In this context we are addressing an important point in the communication with childhood cancer survivors.