Gliomas are the most common primary brain tumours, with glioblastoma multiforme (GBM) being the most frequent, aggressive and invasive tumour type. Postoperative radiotherapy with concomitant temozolomide (TMZ) has become the standard of care for patients with newly diagnosed GBM, based on the results of a large European-Canadian phase III trial [1]. This latter randomised trial demonstrated a significant increase in median survival from 12.1 months after radiotherapy alone to 14.6 months after radiotherapy combined with TMZ. Benefits of TMZ with radiotherapy lasted throughout 5 years of follow-up, with a survival rate of 9.8% versus 1.9% after radiotherapy alone [2]. Despite this important success, most patients die from recurrent disease. The high recurrence rate of about 100% is due to the infiltrative growth characteristics of this tumour type, with its spread throughout normal brain tissue, and high resistance to both radiotherapy and chemotherapy.

Malignant gliomas relapse in up to 90%, in close proximity to the resection site or the initially irradiated volume [3]. Treatment options for recurrent glioma remain limited and include re-resection, chemotherapy, and a second course of radiotherapy. However, there is no standard protocol for re-irradiation of brain tumours. The limited radiation tolerance of normal brain tissue determines the re-irradiation dose that can be applied in addition to the dose of the initial irradiation course, with an acceptable late morbidity profile. A large variety of palliative re-irradiation treatment schemes are reported, with different total dose, number and size of fractions. Retreatment schemes for recurrent gliomas often comprise hypofractionation as well as additional therapy, mostly anti-angiogenic drugs. Generally, the applied re-irradiation technique is chosen based on tumour volume. Only tumour recurrences that are sufficiently small can be treated with high conformality and allow the use of hypofractionated or single dose treatment; this spares normal tissue, which decreases the risk of volume-dependent late toxicity [4,5].

This paper presents an overview of current clinical data on re-irradiation of recurrent glioma with respect to the tolerance dose of normal, healthy brain tissue. To obtain the cumulative radiation dose from the initial and the re-irradiation protocols and to enable comparison of data between studies, rather than taking the “physical” dose, the tolerance dose of normal brain tissue is presented as a ‘biological’ equivalent total dose when applied in 2 Gy fractions (EQD2), estimated by analysis using the linear quadratic model [6,7,8]. Such analysis provides insight into the re-irradiation tolerance of the normal, healthy brain that might be used as a guideline in clinical practice. Particle irradiation has a beneficial dose distribution and should be investigated in the small proportion of patients with small and well described recurrent glioma.

Re-irradiation studies are summarized according to conventional radiotherapy (Table 1), fractionated stereotactic radiotherapy (FSRT) (Table 2) and LINAC-based stereotactic radiosurgery (SRS) (Table 3). Details are provided on histology, time interval between the primary and re-irradiation courses, physical median dose and number of fractions of both irradiation courses, as well as their EQD2 values. Data on patient survival and the probability of side-effects are also shown. Only a few studies are restrictive and investigate a distinct histological subtype, while the other series include a mixture of different histological subtypes with relatively small numbers of patients (range 10–172). With regard to the “treatment volume”, different definitions are used. In most FSRT and SRS series (Table 2 and Table 3), the planning target volume (PTV) is reported, generally including the CT/MR contrast-enhancing tumour with a safety margin ranging from 2 mm to 1 cm.

Data on re-irradiation with conventional external beam radiotherapy are given in Table 1 [9,10,11,12,13,14,15]. The cumulative EQD2 ranged from 81.6 to 102.8 Gy (mean ± S.D: 92.6 ± 6.8 Gy; n = 11) (Table 1). In these series, no information on the PTV was provided. The mean time elapsed between primary radiotherapy and re-irradiation was 29.9 ± 14.1 months (range 14–55 months; n = 7). Acute neurological toxicity (9%) and radionecrosis (6%) were reported in one series of patients treated with a relatively low EQD2_cumulative (87.7 Gy); however, in a twice daily regimen [12]. In contrast, Veninga et al. [13] observed late effects only in those patients receiving a EQD2_cumulative of >102 Gy.

Data on re-irradiation series using LINAC-based FSRT are given in Table 2 [16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]. One exception is the study of Kohshi et al. [24], in which FSRT was delivered with a gamma unit using a noninvasive fixation system.

The cumulative EQD2 ranged from 86.1–133.9 Gy (mean ± S.D: 109.9 ± 13.8 Gy; n = 16) (Table 2). The mean time interval between initial radiotherapy and reirradiation was 16.7 ± 11.1 months (range 3–48 months; n = 17). The mean PTV was 27.6 ± 11.9 cc (range 8.7–51.1 cc; n = 16). Radiochemotherapy with Paclitacel was performed in all 88 patients reported by Lederman et al. [19]. Grosu and coworkers treated 29 of 44 patients with temozolomide [21]. Severe acute radiation-induced toxicity of 8% of the patients was reported in one study (EQD2_cumulative of 99.2 Gy) [17]. In some but not all series, with FSRT, pathologically confirmed radionecrosis was reported in ~2–12% of patients irradiated at EQD2_cumulative > 96 Gy (Table 2). Analysis of 16 studies shows that the incidence of radiation necrosis was not correlated with the EQD2_cumulative or with the time interval between the initial radiation and re-irradiation course.

The LINAC-based SRS re-irradiation data are summarized in Table 3 [17,31,32,33,34,35,36]. The mean time interval between the two courses of radiotherapy was 10.4 ± 1.2 months (range 9.1–12.5 months; n = 6). The mean irradiation volume was 16.9 ± 7.9 cc (range 10–30 cc; n = 7), which is smaller than with conventional re-irradiation or FSRT. The cumulative EQD2 ranged from 111.6 Gy to ~150 Gy (mean ± S.D, 130.5 ± 13.5 Gy; n = 7) (Table 3). Radionecrosis, up to an incidence of 17%, was reported in two studies after an EQD2_cumulative of >137 Gy [17,31]; no additional chemotherapy was administered in these two studies.

Figure 1 shows the influence of the time interval from initial radiotherapy to reirradiation and the EQD2_cumulative on the incidence of radionecrosis. The EQD2_cumulative increased from conventional re-irradiation series (squares) to FSRT (triangles) to SRS series (circles), with a concomitant decrease in the mean time interval between the initial irradiation course and re-irradiation. Despite the higher biological dose and shorter time interval, the incidence of radionecrosis did not differ between the three re-irradiation procedures (Figure 1). A significant correlation (p = 0.016) was found: the higher the EQD2_cumulative, the shorter the time interval between the initial exposure and re-irradiation.

Data on the correlation between the EQD2 of the initial scheme and of the re-irradiation scheme are presented in Figure 2. In patients re-irradiated with conventional radiotherapy, the EQD2_{re-irradiation} was always lower than the EQD2_initial (squares).

In contrast, in FSRT series the EQD2_{re-irradiation} was higher than the EQD2_initial in four out of 16 studies (triangles). One exception (lowest triangle, Figure 2) is the study [24] using re-irradiation in combination with hyperbaric oxygen therapy, resulting in radionecrosis at a relatively low EQD2_{re-irradiation}. In studies using SRS for retreatment of gliomas, the EQD2_{re-irradiation} exceeded the EQD2_initial in all but one of the seven series (circles, Figure 2).

Figure 3 shows correlations between the EQD2_cumulative and the irradiated volume. The figure shows a decrease in treatment volume from FSRT (triangles) re-irradiation series to SRS series (circles). The smaller the re-irradiation volume, the higher the re-irradiation dose applied.

Table 4 presents a summary of the current treatment procedures and treatment data; it shows that increasing conformality of the re-irradiation technique allows exposure to a smaller treatment volume and a higher cumulative biological dose, with a shorter time interval after initial irradiation.

After a comprehensive search (January 1996 to July 2011) 30 brain re-irradiation studies were identified [51]. The keywords included reirradiation, brain tumours, glioma, GBM, external radiotherapy, fractionated stereotactic radiotherapy (FSRT), stereotactic radiosurgery (SRS) and side-effects. Due to the retrospective character of this analysis in most cases only the ‘median’ physical dose and “median” time interval could be considered. Studies in which re-irradiation was combined with chemotherapy are indicated, as well as one study using hyperbaric hyperoxygenation as radiosensitizer [24]. Brachytherapy studies were not included in this analysis.

To enable calculation of the cumulative Equivalent Total Dose (EQD2) only papers with clearly stated median physical dose of the initial radiation treatment are included (Table 1, Table 2, Table 3). The EQD2 represents the total dose if applied in fractions of 2 Gy [7]. The cumulative EQD2 is defined as the sum of the EQD2 of the initial irradiation course and the EQD2 of the re-irradiation course (EQD2_cumulative = EQD2_initial + EQD2_{re-irradiation}). Furthermore, only patients with reported clinically symptomatic necrosis could be considered. It should be realized that severity of symptoms due to focal necrosis is not only based on the size but also on the location of the injury. If the necrotic volume is small and does not include regions like motor cortex or the brain stem, this damage might be unobserved and remain clinically asymptomatic. On the other hand, necrosis of the same size located in one of those sensitive regions can lead to significant morbidity including seizures, symptoms of increased cranial pressure and neuroanatomic-specific symptoms [52,53]. Also, radionecrosis is often difficult to distinguish from tumour recurrence or progression, even when using functional examinations like MRS, PET or SPECT [54]. For details on histopathological and pathophysiologal characteristics of cerebral necrosis, diagnosis and treatments, we refer to Barani and Sneed [53].

In the present analysis, the tolerance dose was defined as the maximum radiation dose that can be tolerated by the normal brain tissue included in the treatment field, or the biological dose that does not induce any irreversible late radiation toxicity. Clinically or histopathologically confirmed brain necrosis, as well as mixed tumour recurrence/necrosis, were considered as “necrosis” due to irradiation beyond the tolerance of the normal brain tissue. Most studies were not designed to measure late toxicity and patients were not actively assessed for neurotoxicity, which might underestimate the incidence of these effects. In Figure 1, only those studies actually reporting on late radiation-induced necrosis are included.

Radiobiological model calculations were performed using the linear quadratic model assuming complete repair between subsequently applied high-dose rate fractions for the clinical studies using conventional radiotherapy (Table 1), FSRT (Table 2), and LINAC-based SRS (Table 3). Treatment regimens were compared using the EQD2 which is based on the Biologically Effective Dose (BED) concept. The BED was calculated with the linear quadratic model [6,7], according to the following formulae: BED = nd (1 + d/[α/β]) [Gy] with d = fraction dose [Gy], n = number of fractions, nd = D = total physical dose [Gy] and the α/β parameter [Gy]. An α/β ratio of 2 Gy was selected for the late responding normal brain [55]. BED values were converted to an Equivalent total dose delivered in 2 Gy fractions (EQD2) using the formula: EQD2 = BED/(1 + d/α/β), which, at fraction size d of 2 Gy and an α/β ratio of 2 Gy = BED/(1 + 2/2) = BED/2. The linear-quadratic model might be less accurate in the high single-dose range, see Discussion (section 3.1). Graphs were prepared using GraphPad (GraphPad Prism 5.01, Software Inc., San Diego, CA, USA). The correlation analysis (Figure 1) was performed using the nonparametric Pearson correlation test, and two-tailed p-values were calculated.

Radiation-induced normal brain tissue necrosis was found to occur at EQD2_cumulative beyond 100 Gy. The applied re-irradiation dose and EQD2_cumulative were found to increase with a change in irradiation technique from conventional to conformal techniques like FSRT, to radiosurgery re-treatment, without increasing the probability of normal brain necrosis. No effect was noticed related to the time interval between the initial and re-irradiation exposure. The mean time interval remarkably decreased with conformality of the re-irradiation therapy. Because of the uniformity of the initial radiation treatment, which is generally a standard regimen of 60 Gy in 2 Gy fractions, our analysis is not conclusive with regard to a likely dependence of tissue recovery on the level of the initial dose.

Finally, in view of the relatively high long-term recovery capacity of the normal human brain, the relatively fast recovery kinetics and tolerance of small treatment volumes to a relatively high total dose, this analysis tends to support further escalation of the target dose in small and well defined recurrences. Because of its beneficial dose distribution, particle irradiation seems to be the superior treatment modality to precisely apply high radiation doses to small target volumes, and should therefore be considered when planning future clinical trials.