In our analysis, 46% of the study patients experienced a local relapse after a median follow-up period of 24 months (range 2–72 months). The majority of local relapses (90%) occurred in the first 5 years after primary therapy including heavy ion-based irradiation, resulting in a 5-year LC rate of 43%, which is in line with other retrospective studies [10, 15, 16]. In contrast to the frequent local recurrences, distant metastases were relatively rare. In only 9% of our study patients metastases were diagnosed in the follow-up imaging after a median follow-up of 42 months (range 12–67 months), with half of these lesions being localized outside the pelvis. Other studies reported higher metastasis rates in the range between 19 and 40% [8, 17–19].
In our analysis, 94% of patients suffered from pain and/or neurological deficits prior to RT. The most serious impairments were observed in patients who had previously had a sacrectomy, including chronic pain, bladder/rectum disorders and urinary and fecal incontinence.
In contrast, the tolerability of the radiation treatment in our study was relatively good. None of the study patients developed urinary or fecal incontinence, severe gastrointestinal toxicity or permanent severe skin damage. In 21% of our study patients, late toxicities ≥ grade 3 were observed affecting the bone and nerve tissues (3 patients with neuropathy, 11 patients with SIFs; see Table 4), and 86% of those patients received a dose of at least 80 Gy (EQD2). Overall, SIFs were by far the most common late toxicity, affecting 49% of the study patients. However, only one third of these fractures (36%) were clinically symptomatic, with severe pain being the main complaint. To this date, there are only 2 other retrospective studies that have specifically investigated the occurrence of sacral fractures after high-dose irradiation of chordomas and reported similar fracture rates [20, 21]. In our analysis, most fractures (85%) occurred within the first 2 years after RT.
During follow-up, 34% of our study patients died. Of these, 74% had a local and/or distant tumor relapse prior to death. For the entire study population, the 5-year PFS, MFS and OS rates were 44%, 82% and 82% after a median follow-up period of 60 months, respectively. In the multivariate analysis we could not identify any significant prognostic factor for LC. However, borderline significance was evident for treatment in the event of locoregional relapse (p-value = 0.07). Irradiation within the primary therapy resulted in 5-year LC rates of 53% compared to 9% when RT was performed in the relapse situation. In line with our results, researchers at the Massachusetts General Hospital (MGH) reported significantly worse LC rates and OS rates when therapy was given in the relapse situation compared to the primary situation [22].
In summary, the main challenge in the treatment of sacral chordoma is to achieve permanent local control after primary treatment. Compared to our results, other particle centers reported better LC rates. The MGH, for example, reported 5-year LC rates of 60% after high-dose, proton-based irradiation with or without surgery after a median follow-up of 41 months [7]. The Nationale Institute of Radiological Sciences (NIRS) in Chiba, Japan could show even better results than the MGH with 5-year LC rates of 77% [8]; in this study, patients received primary hypofractionated heavy ion therapy with a median dose of 70,4 Gy in 16 fractions over 4 weeks. The median follow-up period was 62 months. In total, 29% of local recurrences occurred later than 5 years after irradiation. As in our study, there was no statistically significant prognostic factor for a local relapse in this study.
However, the comparison of previous study results is complicated by the differences in radiation modalities, doses and fractionation, biological models, the populations studied (primary vs. relapse situation, definitive vs. postoperative situation), tumor volumes and target volume definitions, follow-up periods and local relapse definition. In our study 81% of the patients had a macroscopic tumor (median volume 243 ml) and 24% of the patients were irradiated in the local relapse situation. Therefore, our study population represents rather an unfavourable collective compared to many other retrospective studies [8, 23, 24].
Due to the fact that chordomas grow slowly, long follow-up periods are necessary; local recurrences often occur later than 5 years after primary therapy and even after 10 years of follow-up no plateau of local recurrence rates is reached [12]. Therefore, when comparing studies, the follow-up time must be taken into account.
A further problem is that there is no uniform definition of local relapse. In many studies there is no detailed information on this or reference is made only to the radiologist's assessment [7, 8, 24]; Chen et al. defined the local recurrence as an increase in size of chordomas in 2 consecutive follow-up examinations [23]. In another study the modified Response Evaluation Criteria in Solid Tumors (RECIST) and volumetric analysis were used to evaluate the tumor response to RT [13].
Another critical point is the definition of the clinical target volume, which differed significantly between our study and other retrospective studies. In our study the CTV of the basic plan usually included the entire sacrum and in the boost plan this was reduced to a small GTV-CTV safety margin of 3–5 mm. This field shrinkage technique was also used in the MGH studies [7, 22]. In contrast, In the NIRS and HIMBC studies only one CTV was generated for the entire irradiation series by adding a small safety margin to the GTV (as it was done in our study for the boost plan), i.e. often the entire sacrum was not included in the target volume [8, 24]. Consequently, in many other studies sacral tumor recurrences could only have been classified as locoregional recurrences if they were out-of-field or at the edge of the field, whereas in our study they were always classified as local recurrences. Therefore, PFS may be a better comparison parameter than LC when comparing the results of different retrospective series. For instance, LC in the NIRS patient collective was almost twice as high as in our collective (5-year LC 43% vs. 77%), while PFS and OS rates were similar (5-year PFS 44% vs. 50%, 5-year OS 82% vs. 81%) [8]. In this study, univariate analysis found that the patient’s age and the PTV were strong prognostic factors for the OS. However, none of these factors reached statistical significance in the multivariate testing.
A further problem are the different models in Europe and Japan for prescribing the dose of carbon ion therapy, which makes it difficult to compare the doses applied in the various retrospective studies [25]. Therefore, another possible explanation for the better LC rates of the NIRS study could be the higher cumulative equivalent radiation dose applied (64.0–73.6 Gy RBE). However, the radiation dose was no significant predictor for LC in our analysis.
Our study has several limitations. First, the data set was collected retrospectively. The main problem for conducting prospective studies in sacral chordoma is the rarity of the disease, which is why there are only retrospective studies on this topic so far. Secondly, the number of patients in our study is limited, which is also explained by the rarity of the disease; however, there are only a few studies with more patients than in our study [6–8]. Thirdly, the median follow-up time of about 5 years is sufficient for many tumors to collect long-term oncological data, but it is not sufficient for sacral chordomas, as these grow slowly and local recurrences therefore often occur later than 5 years after primary therapy [6]. Therefore, we intend to further observe this patient collective and to republish the collected data again. In addition, a randomized phase 2 study is currently underway at the Heidelberg Ion Therapy Center (ISAC), which compares hypofractionated proton and heavy ion therapy in sacral chordoma patients with a dose of 64 Gy RBE in 16 fractions in both therapy arms; the results are still pending [26].