Background

Primary central nervous system (CNS) lymphoma (PCNSL) is an aggressive extranodal lymphoma exclusively involving the brain, spinal cord, cranial nerves, leptomeninges, and eyes [1]. Diffuse large B-cell lymphoma (DLBCL) is the most common histological subtype of this entity [1]. The incidence of PCNSL is increasing in recent years with a reported age-adjusted incidence of PCNSL of 0.16 per 100,000 [2]. The highest incidence rate was observed among older patients over 65 years. To exclude other lymphoma types with a secondary CNS infiltration, computed tomography (CT) is performed for staging purposes among other diagnostic modalities [1]. PCNSL has different diagnostic implications, treatment regimes, and outcomes compared to lymphoma occurring in the other part of the body, which results in separate observations and studies of this entity [1, 2].

Nowadays, assessment of body composition represents an emergent research field in general medicine, in particular in radiology and oncology [3,4,5,6,7]. So far, the skeletal muscle area and fat areas, most commonly subcutaneous and visceral fat areas, can be quantified on radiological images. Parameters of body composition can predict relevant outcomes in oncology [3,4,5,6,7,8]. For instance, low skeletal muscle mass (LSMM) is an important factor for occurrence of treatment toxicity [8]. However, data regarding body composition in PCNSL are still scarce to date.

Therefore, the aim of the present study was to elucidate the prognostic role of CT-defined body composition parameters on dose-limiting toxicity (DLT) and treatment response in patients with PCNSL.

Methods

Patient acquisition

This retrospective study was approved by the institutional review board (Nr. 145/21, Ethics Committee, Otto-von-Guericke-University Magdeburg, Magdeburg, Germany).

All patients with PCNSL were retrospectively assessed within the time period 2012 to 2020 in one university hospital. Overall, 61 patients (29 female patients, 47.5%) with a mean age of 63.8 ± 12.2 years, range 23–81 years, were identified in the data base with sufficient clinical and imaging data. In every case, the diagnosis of PCNSL was confirmed by histopathological examination after stereotactic biopsy before admission of steroids.

Inclusion criteria for the present study were as follows:

  • -contrast-enhanced staging CT at baseline diagnosis

  • -histopathological diagnosis after biopsy or surgery

  • -first-line standard treatment with radio-chemotherapy

Exclusion criteria were as follows:

  • -no available staging CT or no contrast-enhanced CT

  • -secondary CNS involvement lymphoma

  • -second- or third-line treatment.

Objective response rate

All patients were initially evaluated with magnetic resonance imaging (MRI) of the brain during clinical routine. Response was assessed based on serial brain MRI, beginning within 4 weeks from initiation of induction treatment and reassessed every month thereafter for 4–6 months when possible. Patients then underwent CNS imaging with MRI every 3 months for the first year and followed clinically after stable findings on imaging. No MRIs were performed during steroid therapy. Measurable lesions were required to be at least 2 cm in diameter. Treatment response was defined as complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD). As such, complete response (CR) was defined as complete disappearance of contrast enhancement on MRI. Partial response (PR) is defined as a 50% decrease in enhancing tumor diameters. Progressive disease (PD) is defined as 25% increase in the enhancing lesions and appearance of any new CNS or non-CNS site of disease. Any other situation was characterized as stable disease (SD). Objective response was defined as CR or PR, as previously described [10].

Treatment

Conventional therapy for PCNSL is divided into induction and consolidation phase. In 2020, the National Comprehensive Cancer Network guidelines recommend systemic therapy for patients suitable for or capable of tolerating high-dose chemotherapy, while for unfit patients, 24–36 Gy of whole-brain radiotherapy (WBRT) with a boost to gross disease for a total of 45 Gy is indicated, as first recommendation is a high-dose methotrexate at 8 g/m2 with rituximab and temozolomide or a reduced dose of 3.5 g/m2 methotrexate with rituximab, vincristine, and procarbazine as well as WBRT [1,2,3].

Dose-limiting toxicity

DLT was defined as any treatment-related grade 3 or 4 hepatic toxicity (i.e., aminotransferase levels exceeding 5.1 or 10 times the normal reference values) or any grade ≥ 4 hematologic toxicity (neutrophils < 500/mm3, hemoglobin < 5 g/dL, platelets < 25 × 109/L) during the first treatment course, according to the Radiation Therapy Oncology Group criteria [11]. The DLT was evaluated following 3 months of post-treatment follow-up.

Imaging technique

All CT scans were obtained on a clinical multidetector CT scanner of different vendors (Siemens Somatom Definition AS + , Siemens Healthcare or Canon Aquilion Prime, Canon Medical Systems). The analyzed CT images were obtained before any form of treatment at the baseline staging CT. The CT protocol was as follows: acquisition slice thickness 1 mm with 5-mm reconstructions, tube voltage 120 kV, automatic tube current modulation, pitch factor 1.2, collimation 0.6 mm. In all cases, contrast media was given and the CT scan was acquired in portal venous phase after 40 s after intravenous injection.

All images were assessed in consensus by two experienced radiologists (V.F. and A.S.) who were blinded to the clinical course of the patients. Measurements were performed on axial images in the soft tissue window (window of 45 to 250 HU) on a dedicated workstation (Infinitt PACS, version 3.0, Infinitt Healthcare).

Body composition quantification

Parameters of body composition were semiautomatically measured with the freely available ImageJ software 1.48v (National Institutes of Health Image Program). One axial slice on the mid of the third lumbar vertebra (L3) was used according to the previous descriptions [4, 6]. Skeletal muscle area (SMA) was calculated with the threshold values of − 29 and 150 HU [6]. SMA was divided by the height squared to calculate the skeletal muscle index (SMI). For sarcopenia definition, the SMI threshold proposed by Prado et al was used: 52.4 cm2/m2 for male and 38.5 cm2/m2 for female patients [7].

Fat areas were measured using the HU threshold levels of − 190 and − 30 HU [6]. Visceral adipose tissue (VAT) was measured as the intra-abdominal fat and subcutaneous adipose tissue (SAT) as the fat area located subcutaneously. Visceral to subcutaneous ratio (VSR) was calculated as a ratio of visceral to subcutaneous fat.

The proposed threshold value of 100 cm2 was utilized as a cut-off value to determine visceral and/or subcutaneous obesity [12]. High VSR was defined as 1.1. Figures 1a and b display two representative patients for illustrative purposes.

Fig. 1
figure 1

Representative cases of the patient sample. Red, skeletal muscle area; blue, subcutaneous adipose tissue area; yellow, visceral adipose tissue area; green, intramuscular adipose tissue area. a Case with high skeletal muscle area and normal adipose tissue areas. b Case with high visceral and subcutaneous fat areas and low skeletal muscle area

Full size image

Statistical analysis

The statistical analysis and graphics creation were performed using SPSS (IBM SPSS Statistics for Windows, version 25.0: IBM corporation). Collected data were evaluated by means of descriptive statistics (means, absolute and relative frequencies). Group differences were calculated with the Mann–Whitney-U- test and Fisher exact test, when suitable. Uni- and multivariable regression analysis was used to elucidate possible associations. In all instances, p values  were  interpretated exploratorely.

Results

The patients received in median 4 cycles of chemotherapy, range 1–8 cycles.

Overall, 19 patients (31.1%) had progressive disease, 23 patients showed complete response (37.7% of all patients and 65.7% of patients with treatment response), and 12 patients had partial response (19.7% of all patients and 34.2% of patients with treatment response), and 3 patients had stable disease (4.9%). For 4 patients (6.6%), no treatment was possible, and therefore no follow-up was performed.

The mean values of the analyzed body composition parameters were as follows: VAT, 165 ± 101.5 cm2; SAT, 190.8 ± 82.5 cm2; VSR, 0.96 ± 0.71; SMI, 45.8 ± 9.7 cm2/m2. Visceral obesity was identified in 35 patients (57.4%).

Associations between body composition and DLT

DLT occurred in 28 patients (45.9%). No body composition parameter was associated with occurrence of DLT (Tables 1 and 2).

Table 1 Body composition parameters in patients with and without DLT
Full size table
Table 2 DLT and number of cycles according to the dichotomized body composition parameters
Full size table

However, patients with normal VSR could receive more chemotherapy cycles compared to patients with high VSR (mean 4.25 vs 2.94, p = 0.03). There were no strong associations between dichotomized body composition parameters and metric parameters on DLT (Table 3).

Table 3 Influence of body composition parameters on DLT (univariable analysis)
Full size table

Associations between body composition and ORR

Patients with objective response had higher muscle density values compared to patients with SD and/or PD (Table 4). Moreover, muscle gauge was also higher in patients with objective response. Other body composition parameters were not associated with objective response.

Table 4 Body composition parameters in patients with and without objective response
Full size table

Regression analysis revealed that sarcopenia predicted OR, OR = 5.19 (95% CI 1.35–19.94) p = 0.02 (univariable regression), and OR = 4.23 (95% CI 1.03, 17.38), p = 0.046 (multivariable regression) (Table 5).

Table 5 Influence of body composition parameters on objective response
Full size table

Discussion

The present study used CT-defined body composition parameters for prediction DLT and treatment response in patients with PCNSL. As a key finding, body composition parameters were not able to predict DLT but sarcopenia was strongly associated with objective response. Therefore, assessment of body composition in PCNSL can be recommended for prediction of treatment course.

The emerging field of body composition utilized cross-sectional imaging modalities acquired for diagnostic purposes and can provide novel quantitative biomarkers of the constitution of the body [3,4,5,6,7,8,9]. A plethora of studies were published investigating the prognostic relevance of LSMM and adipose tissue throughout medicine, predominantly in oncology [3,4,5,6,7,8,9].

Notably, most studies tested the prognostic implications of body composition parameters but not their predictive role [3]. However, for direct treatment guidance, the predictive role of an imaging biomarker is of great importance. As a novel finding of the present study, sarcopenia is strongly associated with objective response. There are no published data regarding possible associations between body composition and objective response.

According to the literature, established independent prognostic factors in PNCSL are age, Karnofsky performance index, sex, and response to induction chemotherapy [13]. Regarding predictive imaging markers, a recent study could highlight the important role of dynamic contrast-enhanced MRI, which was correlated with objective response [14]. However, assessment of sarcopenia is a by-product of existing imaging data, whereas dynamic contrast-enhanced MRI must be obtained additionally with a further time consumption.

Several reasons could cause the identified association between sarcopenia and objective response. It was clearly shown that skeletal muscle area as a surrogate parameter for overall body constitution is an important factor for chemotherapy tolerance and effectiveness. As such, in a recent meta-analysis, sarcopenia was associated with overall therapy toxicity, OR = 2.19, 95% CI 1.78–2.68 [8]. This was clearly shown as well as in curative as palliative setting. Furthermore, the present study identified that also muscle quality expressed by muscle density and muscle gauge was associated with objective response. Finally, an important finding of the present study is that VSR might be an interesting factor to predict the number of chemotherapy cycles in patients with PCNSL.

Previously, only few studies analyzed the prognostic role of body composition in PCNSL. In a recent study, it has been shown that LSMM/sarcopenia predicted progression-free survival (HR = 4.40, 95% CI 1.66–11.61, p = 0.003) and overall survival (HR = 3.16, 95% CI 1.09–9.11, p = 0.034) [15]. Similar results were reported also by other authors [16].

According to the literature, measure of the muscle status on CT at the L3 level represents a standardized method to quantify the skeletal musculature [17]. The employed method, also used in the present study, provides reliable and validated results [6]. In PCNSL, previous investigations used different measurements, levels, and values for estimation of sarcopenia. So far, Leone et al measured skeletal muscle status on the L3 level as well as temporal muscle thickness derived by brain MRI [15]. Furtner et al used only the temporal muscle thickness [16]. It was shown that temporal muscle thickness was an independent predictor of mortality (HR = 2.5, 95% CI 1.6–3.9, p < 0.001) [16]. Clearly, further studies are needed to compare the prognostic and predictive role of temporal muscle thickness as a surrogate parameter of muscle status in comparison to the standardized approach on the level of L3. Importantly, every patient with PCNSL needs a staging CT to rule out extracranial manifestations [2]; analysis of body composition can be performed in clinical routine without the need for new scans.

Interestingly, as reported previously, reduced muscle quality and increased intermuscular fat were associated with poor prognosis in DLBCL [18]. However, in the present study, no associations between DLT and/or objective response and intramuscular adipose tissue could be found.

The present study is not free from limitations. First, it is a retrospective analysis of one center. This fact might result in a selection bias. Second, the time delay between CT imaging and treatment differed slightly due to clinical routine. However, the effect of treatment on body composition might be neglected in a short time frame. Third, treatment regime differed between the patients. However, this reflects the daily clinical routine. Fourth, contrast media phase could have an influence on HU measurement of the muscle and fat density. However, all patients were investigated in portal venous phase, which should reduce possible bias. Beyond that, the absolute HU differences between different contrast media phases are relatively small [19].

In conclusion, sarcopenia is strongly associated with objective response in patients with PCNSL. None of body composition parameters can predict DLT in PCNSL.