Prognostic value of tumour volume in patients with a poor Karnofsky performance status scale – a bicentric retrospective study

Backround Median overall survival (OS) after diagnosis of glioblastoma (GBM) remains 15 months amongst patients receiving aggressive surgical resection, chemotherapy and irradiation. Treatment of patients with a poor preoperative Karnofsky Performance Status Scale (KPSS) is still controversial. Therefore, we retrospectively assessed the outcome after surgical treatment in patients with a KPSS of ≤60%. Methods We retrospectively included patients with a de-novo glioblastoma WHO °IV and preoperative KPSS ≤60%, who underwent surgery at two neurosurgical centres between September 2006 and March 2016. We recorded pre- and postoperative tumour volume, pre- and postoperative KPSS, OS, age and MGMT promoter status. Results One hundred twenty-three patients (58 females/65 males, mean age 67.4 ± 13.4 years) met the inclusion criteria. Seventy-five of the 123 patients (61%) underwent surgical resection. 48/123 patients (39%) received a biopsy. The median preoperative and postoperative tumour volume of all patients was 33.0 ± 31.3 cm3 (IR 15.0–56.5cm3) and 3.1 ± 23.8 cm3 (IR 0.2–15.0 cm3), respectively. The median KPSS was 60% (range 20–60%) preoperatively and 50% (range 0–80%) postoperatively. Patients who received a biopsy showed a median OS for patients who received a biopsy only was 3.0 months (95% CI 2.0–4.0 months), compared to patients who had a resection and had a median OS of 8 months (95% CI 3.1–12.9 months). Age (p < 0.001, HR: 1.045 [95% CI 1.022–1.068]), postoperative tumour volume (p = 0.02, HR: 1.016 [95% CI 1.002–1.029]) and MGMT promotor status (p = 0.016, HR: 0.473 [95% CI 0.257–0.871]) were statistically significant in multivariate analysis. In subgroup analyses only age was shown as a significant prognostic factor in multivariate analyses for patients receiving surgery (p < 0.001, HR: 1.046 [95% CI 1.022–1.072]). In the biopsy group no significant prognostic factors were shown in multivariate analysis. Conclusion GBM patients with a preoperative KPSS of ≤60% might profit from surgical reduction of tumour burden.

with systemic malignancies and divided them according to their level of activity and medical requirements. Patients are scored into 11 categories from 0 to 100, where, for example, a KPSS of 70% means the patient is able to care for himself but is unable to carry out daily activities [1]. After it had been proven successful in patients with systemic cancer, more and more research groups started to evaluate the KPSS score for brain cancer [2][3][4]. Previously published studies could show a significant correlation between the preoperative KPSS score and the outcome after glioma surgery [5,6]. In most studies, only patients suffering from a glioblastoma with a KPSS of ≥70% were included [7,8]. For example, those studies analysed prognostic factors such as tumour size, GTR and adjuvant therapy modalities postoperatively. However, in our clinical daily work, patients with a noticeably lower KPSS are represented as well. It should be noticed that this can be due to clinic symptomology as seizures, acute mental status changes or focal neurologic deficits caused by tumour size and/or location itself. Therefore, the following study intends to show whether it is worthwhile for patients with a KPSS 60% or below to achieve tumour volume reduction.

Methods
This retrospective, non-interventional bicentric study was approved by the medical ethics committee of the Technical University Munich (5625-12) and is in accordance with the ethical standards of the 1964 Declaration of Helsinki and its later amendments [9].

Patient population
We retrospectively assessed 968 patients with a histologically confirmed glioblastoma WHO IV with a preoperative Karnofsky Performance Status Scale (KPSS) of ≤60%, who were treated surgically between September 2006 and March 2016 in two neurosurgical departments (Fig. 1). According to interdisciplinary neuro-oncological consensus, patients were assigned to surgery with the intent of complete resection or to biopsy to confirm the histopathological diagnosis.
We retrospectively reviewed pre-and postoperative KPSS, date of initial tumour diagnosis, date of death/ last contact, age, sex, adjuvant treatment and histopathological findings from the patients' medical charts. Also, we performed histopathological analysis according to the WHO criteria of 2016 [10] and quantitatively assessed methylation of the O6-methylguanin-DNAmethyltransferase (MGMT) promoter status. Since the decision regarding adjuvant therapy is made after receipt of the histology in the context of the interdisciplinary neuro-oncological board, depending on the clinical condition, the KPSS was collected approximately 5 days postoperatively.
Then, we calculated the overall survival (OS) from the date of surgery until the date of death or censored for the date of the last patient contact. Only patients with complete magnetic resonance imaging data were included to calculate pre-and postoperative contrast-enhancing tumour volumes. Patients with recurrent tumour or incomplete data were excluded (Table 1).

Imaging
All patients received preoperative and early postoperative MRI (within 72 h after surgery). In centre A, we performed imaging using three different 3 Tesla MRI scanners: Philips Achieva; Philips Ingenia (Philips Medical Systems, The Netherlands B.V.); and Siemens Verio (Siemens Healthcare, Erlangen, Germany). Images included T1w sequences with and without contrast agent, FLAIR (Fluid attenuated inversion recovery) sequences, T2 gradient echo sequences, diffusion-weighted imaging or diffusion-tensor imaging, whereas we calculated isotropic diffusion-weighted images and apparent diffusion coefficient (ADC) maps automatically. Tumour volumes of the contrast-enhancing tumour on pre-and early postoperative MR images using iPlannet ® Cranial 3.0.1 were manually segmented by two neurosurgeons (5 and 10 years of experience) and two neuroradiologists (3 years and 6 years of experience).
In centre B, we conducted MR imaging with a 3.0 T MRI scanner (Biograph mMR, Siemens Healthcare, Erlangen, Germany). One neurosurgeon (14 years of experience) and one medical student assessed the volumes of the contrast-enhancing tumour through manual segmentation via iPlannet ® Cranial 3.0.1 (iPlannet ® 3.0 cranial planning software, Brainlab AG, Munich, Germany). The postoperative tumour volumes of patients who underwent biopsies were considered identical to the preoperative tumour volumes.

Statistical evaluation
We conducted our data analysis using IBM SPSS Statistics Version 24.0 and 26.0 (SPSS Inc., IBM Corp., Armonk, NY, USA). In the descriptive data analysis, we show non-normally distributed data as median and interquartile range (IR), normally distributed variables as mean and standard deviation.
We compared the OS distributions using the Kaplan-Meier estimates (log-rank) and a Cox regression model for multivariate survival analysis. We considered differences with an error probability of less than 0.05 to be statistically significant.

Patients and clinical data
123/968 patients (58 females/65 males) with a mean age of 67.4 ± 13.4 years; (range 21-90 years) met our inclusion criteria: surgical treatment for glioblastoma, preoperative KPSS of ≤60%, preoperative and early postoperative MRI, complete medical documentations with date of initial tumour diagnosis, date of death/last contact, age, sex, adjuvant treatment and histopathological findings. Data are shown for all patients and for the subgroups of patients with biopsy / surgery ( Table 1).
Surgical resection with intent for maximum/complete resection was performed in 75/123 patients  Assuming that the adjuvant therapy could be started 14 days after the operation and lasted approximately 6 weeks, we could suppose the completion of the adjuvant therapy in all but nine patients on the basis of the OS. Unfortunately, the follow-up expired without documentation regarding this information.

Overall survival (OS)
Median OS was 5.0 months (95% CI 3.0-4.0 months) in the study population (including biopsy group and surgical resection group). Patients who received a biopsy showed a median OS of 3.0 months (95% CI 2.0-4.0 months), whereas patients who underwent surgical resection showed a median OS of 8.0 months (95% CI 3.1-12.9 months).
In order to show a possible survival advantage of the resected patients compared to the biopsied patients, we performed a Cox regression with the parameters: age, postoperative tumour volume, preoperative KPSS and biopsy. This showed no more significance for the biopsied patient group (p = 0.154, 95% CI 0.399-1.156).
At the time of the study, 102/123 patients (82.9%) had died, and 21/123 (17.1%) were still alive or censored for their last date of contact. In-hospital mortality was seen

Univariate model
Surgical resection compared to biopsy (p < 0.001) and complete resection of the contrast-enhancing tumor part (p = 0.032) showed a significant impact on OS in the univariate analysis using Kaplan-Meier estimates. MGMTmethylation status did not show a significant impact on OS in univariate analysis (p = 0.071) ( Figs. 2A-B and 3). Adjuvant therapy regimes also showed a significant prognostic impact, in all patients (p < 0.001) and in the subgroups of patients with biopsy (p = 0.005) and surgery (p < 0.001) (Fig. 4A-C).

Multivariate model
Cox

Discussion
In this cohort of GBM patients (biopsy group and surgical resection group) with a preoperative KPSS ≤60%, postoperative tumour volume, age at the time of surgery and MGMT-methylation status were significant predictors of OS in the multivariate analysis. In  Fig. 4 Overall survival, categorized in STUPP regime/RTX alone: A) Complete cohort B) Resection C) Biopsy contrast, preoperative tumour volume and KPSS had no significant impact on OS. The subgroup analysis of patients referred for surgery only age was shown as prognostic factor in multivariate analysis, whereas no significant prognostic factors were shown in the subgroups of patients referred for biopsy. In our study population, the significant effect of postoperative tumour volume in univariate analysis correlates with tumour resection compared to the biopsied group. This could explain the lack of significance in the multivariate analysis for this group. Nevertheless, as already understood from other studies, we could also show that extent of resection is an important factor in OS in patients with glioblastoma [11][12][13][14].
In general, patients with poor preoperative KPSS usually do not receive aggressive surgical therapy. Therefore, data on these patients are very limited [15]. In our cohort, 56/123 (45.5%) showed an improved or unchanged postoperative KPSS with a median of 50%. Adjuvant treatment such as radiation therapy or chemotherapy is usually only offered to patients with a KPSS ≥70% [16,17]. Consequently, these patients are usually considered ineligible for adjuvant oncological treatment even after tumour resection. Malakhov et al. could show that 51.2% of the patients presenting with KPSS< 60% and receiving chemoradiation had improved survival compared to RT alone [18]. However, the majority of our patient cohort (77.6%) who underwent surgical resection received adjuvant therapy. Considering the early postoperative assessment of KPSS in this study, secondary improvement is to be expected. Patients undergoing a biopsy were older (66.6%) with eloquent tumour location than patients, who were selected for surgical tumour resection (41.33%). Only 16.7% of the patients who received a biopsy underwent adjuvant treatment regimes. On the one hand, this is due to the higher age of the biopsy group. Secondly, only 7/48 of these patients showed MGMT-methylation. In accordance with the guidelines, our neuro-oncological interdisciplinary board recommends monotherapy for patients ≥75 years of age, depending on the MGMTmethylation status.
Reduced preoperative KPSS is an important prognostic factor in patients with glioblastoma [19,20]. Age, comorbidities and neurological deficits have an impact on KPSS and, in conclusion, on OS [20][21][22]. Postoperative deterioration of the performance status scale is usually multifactorial, with the reasons being edema, haemorrhage, postoperative delirium, ischemic events or direct surgical lesions of eloquent brain structures [23].
In our opinion, the KPSS does not offer sufficient information about quality of life and therefore should not be overrated concerning the selection of patients undergoing surgery. For example, patients with preoperative neurological deficits such as hemiparesis due to surrounding edema might have a KPSS of 60% or below and might therefore not be selected for surgical therapy. However, as we know today, the surrounding edema will disappear a few days after surgery, and the patients are able to recover for adjuvant treatment. The KPSS should therefore be considered with care.
The decision for or against aggressive surgical therapy should be made individually by experienced neurosurgeons within the framework of an interdisciplinary neuro-oncology board.

Limitations of the study
This study has limitations. First, the retrospective nonrandomized design is the main limitation. Second, molecular status was not available for all patients in our cohort study, as the MGMT-methylation status of patients with glioblastoma is known to be one of the strongest predictors concerning survival prognosis [24,25].

Conclusion
GBM patients with a preoperative KPSS of ≤60% might profit from surgical reduction of tumour burden. We therefore suggest considering surgical resection even in patients with a KPSS of ≤60% after careful selection based on an interdisciplinary neuro-oncological board decision and counselling of patients and their relatives.