In this retrospective population-based cohort study we found that 9.4% of adults diagnosed with glioblastoma met the criteria for butterfly glioblastoma. Outcome was poor, with a median overall survival of less than six months. This study adds real-world data on prognostic factors and outcome in a previously sparsely described subgroup of glioblastoma, whereas previous studies are based on data from tertiary referral centres [3-5, 8].
Survival
Our study confirms previous findings of a considerably poorer prognosis in butterfly glioblastoma compared to glioblastoma in general, where population-based studies observed a median overall survival of nine to 11 months [22-25]. A median overall survival of 5.9 months in our cohort was comparable to the results from previous butterfly glioblastoma studies, with median survival ranging from 3.2 to 5.9 months [4, 5, 8]. Survival was substantially poorer than that observed by Burks and colleagues, where median survival was 12 and 15 months in patients who underwent two different surgical approaches [6]. Likely explanations are different selection criteria and settings, and highly different resection rates of 16% and 100%. Although these two studies cannot be compared directly, this may indicate a potential benefit from a more aggressive treatment approach in patients with butterfly glioblastoma. Patients aged over 70 years received less intensive treatment than younger patients, and had a significantly poorer outcome with a median survival of only two months, equal to the findings of Dayani and colleagues [4]. Previous studies have demonstrated that combined chemoradiotherapy improves outcome in elderly glioblastoma patients with adequate performance status [26-28]. Future studies, preferably including quality of life analyses, may clarify if these results are applicable to the butterfly glioblastoma subgroup.
Despite the poor overall outcome, patients with long-term survival existed. Three- and five-year survival rates of 8.3% and 5.6% were comparable to the corresponding pooled survival rates of 11% and 4% in a large meta-analysis of glioblastoma in general [29]. We suggest that patients with butterfly glioblastoma might receive less-intensive treatment due to the deep-seated tumor location, uncertainty of treatment benefit, and the dismal prognosis. This study demonstrates that long-term survival is possible, supporting an argument for a higher treatment intensity in patients in acceptable general condition. A second argument for providing more multimodal treatment is the challenge in predicting treatment benefit, long-term survival, and quality of life in these patients. Only a minority of patients receiving TMZ suffered from significant toxicity, serving as a third argument to promote an increased treatment intensity in patients with butterfly glioblastoma.
Approximately 40% of the patients diagnosed with recurrent glioblastoma received anti-tumor treatment, and had a median post-recurrence survival of nearly eight months. We consider this unexpectedly long, and suggest that anti-tumor treatment may be appropriate in patients with recurrent butterfly glioblastoma and acceptable performance status.
MR volumetric analyses
Correlation analyses revealed a slight negative correlation between larger abnormal T2/FLAIR volumes and age, but without statistical significance. All four patients with the largest tumor volumes in T1-weighted MRI, and four of five patients with the largest abnormal T2/FLAIR volumes, were aged under 70 years. Among possible explanations, we suggest that younger patients, having a higher cognitive reserve, tolerate larger tumor growth before symptoms occur. In addition, IDH mutated glioblastoma is more frequent in younger patients, and is associated with less aggressive growth and larger tumor volumes, although this was not evaluable in this small sample [30].
Adjusted Cox regression revealed that larger abnormal T2/FLAIR volume was slightly negatively associated with survival. This negative prognostic value appears after adjustment for age and sex, most likely explained by higher age being a negative confounding variable. Younger age was strongly associated with longer survival, possibly masking the negative effect of larger T2/FLAIR abnormality in the unadjusted model. Nevertheless, the tumor volumes of long-term surviving patients differed widely, ranging from the upper to lower quartiles. We suggest that larger T2/FLAIR abnormality cannot be considered an unequivocal negative prognostic factor, as larger volumes may be seen in younger patients harbouring favourable prognostic factors. The association between T2/FLAIR volumes and age has not been studied previously, thus comparison with other studies was not applicable. Radiation therapy is a key modality in the treatment of patients with glioblastoma, but abnormal T2/FLAIR volume is not necessarily fully included in the Clinical Target Volume [31, 32]. Future studies should explore any associations between survival and the proportion of T2/FLAIR volume included in the target volume.
In our cohort, both median T1 and T2/FLAIR volumes were larger in the resection group than the non-resection group. Larger tumor volumes in younger patients and larger tumor volumes in frontal lobes, with favourable surgical accessibility, are possible explanations. In contrast, Dziurzynski and colleagues reported slightly larger, but statistically insignificant, median T1 volume in the non-resection group compared to the resection group, although ranges were wide in both groups and with no difference in FLAIR volumes [5]. Among possible explanations for these differences are the different settings, small sample sizes, low resection rates of 16% and 36%, and the lack of adjustment for molecular subgroups. Unlike Dayani and colleagues, who reported a slight negative correlation between larger T1 volume and survival, we found no such association [4]. Small sample sizes and wide range of tumor volumes may explain this uncertainty.
Strengths and limitations
Strengths of this study included the population-based study design and that all patients were followed up for seven years or until death. Furthermore, clinical information on all patients was available in a common electronic record, and T1-weighted and T2-weighted/FLAIR images were available in all patients. The small sample was the main limitation, related to the rarity of the condition, but is comparable to sample sizes in previous studies [3-5, 8]. Other limitations included the lack of histological samples in approximately 50% of the enrolled patients, causing a risk of inclusion bias, and the high number of missing molecular data. Further, associations between treatment and outcome in elderly patients were inconclusive, as only a minority of elderly patients received multimodal treatment.
To conclude, outcome from butterfly glioblastoma was dismal, with a median overall survival of less than six months. However, long-term survival was comparable to that observed in glioblastoma in general, and multimodal treatment was associated with improved survival. This suggests that patients with butterfly glioblastoma may benefit from a more aggressive treatment approach despite the overall poor prognosis.