Ultra-early response assessment in lymphoma treatment: [¹⁸F]FDG PET/MR captures changes in glucose metabolism and cell density within the first 72 hours of treatment

Positron emission tomography (PET) using the radiotracer [¹⁸F]FDG (2-[¹⁸F]fluoro-2-deoxy-d-glucose) is the current imaging technique of choice to assess treatment response in the majority of lymphoma subtypes [1‐5]. It allows direct assessment of the glycolytic activity at the cellular level, which correlates with cell proliferation [6]. Diffusion-weighted imaging (DWI), an advanced magnetic resonance imaging (MRI) technique, can be used to assess tissue diffusivity which is an indirect measure of cell density. Since extracellular water movement in tumours is affected by cell density [7], DWI has recently been suggested as a potent alternative to [¹⁸F]FDG PET/CT to assess treatment response in lymphoma patients [8‐12].

Previous studies have demonstrated that a reduction in glycolytic activity (i.e. [¹⁸F]FDG uptake on PET), and an increase in diffusivity (i.e. reduction of diffusion restriction on DWI, chiefly reflecting a reduction in cell density) can be observed after only a small number of therapy cycles [10, 12‐15]. Notably, interim PET, which is typically performed after two to four therapy cycles and is included in the Lugano classification of the International Conference on Malignant Lymphoma (ICML) [1, 13], is currently under evaluation as a prognostic marker for different lymphoma subtypes, treatments and time-points, but so far with mixed results [15‐22]. However, little is known about in vivo changes at the onset of therapy – in particular, to what extent do changes in glycolytic activity and cell density correlate with each other, and which method detects the earliest changes. In a murine lymphoma model, it has been shown that [¹⁸F]FDG PET can capture changes after treatment with etoposide as early as 16 h after treatment initiation [23]. So far, no comparable in vivo data in lymphoma patients are available. DWI has been shown to capture treatment response in lymphoma in vivo as early as 1 week after therapy initiation, but no earlier time-points were evaluated [24]. Finally, in a study in patients with Hodgkin lymphoma (HL) and diffuse large B-cell lymphoma (DLBCL), no significant changes in apparent diffusion coefficients (ADC) were observed approximately 2 days after therapy initiation, whereas [¹⁸F]FDG PET revealed treatment responses after 19 days [25].

Here, we investigated whether standard chemotherapy or immunochemotherapy led to significant changes in glucose metabolism, cell density, and lesion volume in lymphoma patients, at two time-points during the first week after treatment initiation. We particularly aimed to determine whether such ultra-early effects of treatment on glycolytic activity and cell density of lymphomas occur simultaneously, and whether they differ among histological subtypes, therapeutic regimens, and individuals with the same histology and treatment.

Patients with histologically proven, previously untreated malignant lymphoma who were referred to our institution for routine pretherapy [¹⁸F]FDG PET/MR were eligible for participation in our prospective exploratory study. Ethics committee approval and written informed consent from all patients were obtained prior to imaging. Inclusion criteria were the presence of one of the following histological lymphoma subtypes, as verified by a reference pathologist who analysed tissue samples (obtained by biopsy or during surgery) according to the current WHO classification of haematological and lymphoid malignancies, in combination with the following specified treatment schemes:

Patients with any of the following were excluded: women with a positive pregnancy test or clinically confirmed pregnancy; inability to understand the study goals or outline, or to consent to participation; age below the specified minimum of 18 years; known contraindication to MRI (e.g. implantable medical device according to the MRI Safety Guidelines, or conditions such as claustrophobia); or a blood glucose level >150 mg/dL. Patients who fulfilled the above requirements for participation underwent [¹⁸F]FDG PET/MR at three time-points:

Of 65 patients who fulfilled the inclusion criteria, seven were excluded at baseline due to elevated blood glucose levels (>150 mg/dL). Thus, a total of 58 patients (28 women and 30 men; mean age 58.9 ± 17.6 years, range, 25–92 years) fulfilled the criteria for participation in the study. Of these 58 patients, 11 were diagnosed with HL, 25 with DLBCL (15 in the R-CHOP arm, and 10 in the DA-EPOCH-R arm), 12 with FL, and 10 with MCL. One patient with FL and one patient with MCL underwent [¹⁸F]FDG PET/MR at baseline and FU-1, but not at FU-2; whereas one patient with DLBCL (R-CHOP arm) underwent [¹⁸F]FDG PET/MR at baseline and FU-2, but not at FU-1. A total of 166 target lesions (134 nodal, 32 extranodal; Table 1) were analysed.

On a per-lesion basis, pooled SUVmax, SUVmean, ADCmin and ADCmean values from all lymphoma subtypes differed significantly between baseline and FU-1 (−46.8%, −33.3%, +20.3% and +14%, respectively; P < 0.001), between baseline and FU-2 (−65.1%, −49%, +50.7% and +32.4%, respectively; P < 0.001), and between FU-1 and FU-2 (−34.5%, −23.5%, +24.2% and +15.2%, respectively; P < 0.001). MTV also differed significantly between baseline and FU-1 (−46%; P = 0.002), and between baseline and FU-2 (−61.1%; P < 0.001), but not between FU-1 and FU-2 (−28%; P = 0.79) whereas VOL differed significantly only between baseline and FU-2 (−24.2%; P < 0.001), and not between baseline and FU-1 (−12.8%; P = 0.079), or between FU-1 and FU-2 (−13.1%; P = 0.154; Table 2, Figs. 1 and 2).

On a per-patient basis, pooled SUVmax, SUVmean, ADCmin and ADCmean values from all lymphoma subtypes differed significantly between baseline and FU-1 (−54.9%, −41.7%, +27.8% and +17.8%, respectively; P < 0.001), and between baseline and FU-2 (−70.7%, −55.9%, +58% and +34.4%, respectively; P < 0.001). SUVmean, ADCmin and ADCmean, but not SUVmax, differed significantly between FU-1 and FU-2 (−24.6%, +23.6%, +14.1% and −35%, respectively; P = 0.015, P < 0.001, P < 0.001 and P = 0.094). MTV also differed significantly between baseline and FU-1 (−42.9%; P = 0.01), and between baseline and FU-2 (−53.9%; P = 0.001), but not between FU-1 and FU-2 (−19.2%; P = 1.0), whereas VOL differed significantly only between baseline and FU-2 (−18.7%; P = 0.002), and not between baseline and FU-1 (−12%; P = 0.069), or between FU-1 and FU-2 (−7.6%; P = 0.62). Line graphs revealed major differences in treatment response between individual patients, and even between patients with the same lymphoma subtype and treatment regimen (Online Resource 1).

In the DLBCL group, all patient-based imaging parameter changes from baseline to FU-1 except for ADCmin differed significantly between the two treatment arms (R-CHOP vs. DA-EPOCH-R), whereas only changes in SUVmean and VOL between baseline and FU-2 differed significantly between the two treatment arms (Table 3, Online Resource 2).

Weak, statistically significant negative correlations were observed between lesion-based ΔSUVmax and ΔADCmin (baseline to FU-1 r = −0.32, P < 0.001; baseline to FU-2 r = −0.32, P < 0.001), whereas moderate significant negative correlations were observed between ΔSUVmean and ΔADCmean (baseline to FU-1 r = −0.54, P < 0.001; baseline to FU-2 r = −0.41, P < 0.001).

The observed decrease in lymphoma glucose metabolism (i.e. FDG uptake, measured in terms of SUV and MTV) and increase in tissue diffusivity (i.e. ADC, chiefly reflecting cell density) in our study clearly suggest the presence of treatment effects, which can be measured in vivo by [¹⁸F]FDG PET/MR as early as 48–72 h after treatment initiation. Further imaging parameter changes reflecting treatment effects clearly occurred during the first week. Notably, significant changes during the first week after treatment initiation were found not only in SUVs (and the associated MTVs) and ADCs, but also in morphological volumes, albeit to a lesser degree.

A closer look at the rates of change of the imaging parameters clearly suggests that, regardless of histological subtype, glucose metabolism-based measures (SUV and MTV) show a more pronounced response to treatment during the first 48–72 h after treatment initiation, whereas the changes between 72 h and 1 week appeared to be less pronounced. For diffusivity/cell density, as reflected by ADCs, the opposite effect was generally observed, with a less pronounced response to treatment during the first 48–72 h after treatment initiation, and a more pronounced response between 72 h and 1 week; only DLBCL, in which changes in ADC over time more closely resembled those in SUV, differed in this regard. Significant changes in morphological lesion volumes were also observed at these early time-points, although they were only small-to-moderate, particularly during the first 48–72 h after treatment initiation, when, contrary to changes after 1 week, they did not reach statistical significance. This very early reduction in lesion size under treatment confirms the results of Horger et al. [24], who found a significant reduction in lesion size in a small sample of responding lymphoma patients 1 week after treatment initiation. The fact that, in the latter study, lesion shrinkage was less pronounced than in the present study may possibly be explained by the fact that transaxial lesion diameters, instead of volumes, were used, and also the study population of Horger et al. included only three patients with DLBCL, which was the subtype that showed by far the largest volume decreases in our study.

Clear differences in treatment responses and associated changes in imaging parameters were seen among the four histological lymphoma subtypes in our study. DLBCL showed the largest changes, followed by HL, followed by MCL, with FL showing the smallest changes. On the other hand, only patients with FL and MCL received the same treatment regimen, whereas treatment regimens differed among those with DLBCL and HL, and hence the latter two subtypes cannot be directly compared with each other, or with FL and MCL. Thus, while it is striking that the different degrees of treatment response among the four histological subtypes appear to reflect the order of aggressiveness of these lymphomas, the main finding on this point is that both glucose metabolism and cell density of the lesions show a marked response to their respective standard therapy regimens in all examined lymphoma subtypes as early as 48–72 h after treatment initiation.

Notably, in DLBCL, for which two different treatment regimens (R-CHOP and DA-EPOCH-R) were used, the observed differences between the two treatment groups point towards a stronger and/or more rapid effect of R-CHOP on glucose metabolism and diffusivity/cell density of the lymphomas in this early phase, which seems plausible because, with R-CHOP, both rituximab and the chemotherapeutic agents were administered on day 1, whereas with DA-EPOCH-R, rituximab was administered on day 1, and the chemotherapeutic agents were administered over 5 days. Thus, at the first PET/MR follow-up at 48–72 h after treatment initiation, patients in the DA-EPOCH-R arm had only received a fraction of the total chemotherapy dose, and lower peak levels. However, assignment to the two treatment groups was based on IPI score or the presence of double-hit lymphomas, i.e. only high-risk patients were chosen for treatment with DA-EPOCH-R, in accordance with previous studies [29, 30]. As a consequence, it is quite possible that the clinical entities in themselves are not comparable, and this led to, or at least contributed to, the observed differences in ultra-early responses on [¹⁸F]FDG PET/MR.

As mentioned above, DWI enables (indirect) assessment of cell density through assessment of tissue diffusivity, based on the underlying mechanism that in hypercellular tumours such as lymphoma, the extracellular space is compressed, limiting the local Brownian motion of water molecules. As a consequence, DWI may be used to capture treatment effects in terms of cell necrosis [31]. It is highly unlikely that, in our study, an isolated treatment-induced inflammatory fluid influx (which might have led to a T2-shine-through effect), rather than cell necrosis, was responsible for the observed ADC increase after treatment for several reasons. Firstly, a fluid influx in the absence of cell necrosis would lead to an increase in lesion size, which we did not observe. On the contrary, there was a small-to-moderate statistically significant decrease in morphological lesion volumes, which is indicative of cell necrosis. Secondly, a possible treatment-induced inflammatory reaction would not explain the obvious differences in ADC increase among the four histological lymphoma subtypes, as this would require a histology-dependent or treatment-dependent degree of inflammation for which there is presently no evidence; even more so as three of the lymphoma subgroups received rituximab-based treatment. Here, it is also important to note that treatment regimens for HL and DLBCL included the corticosteroid prednisone, which has been shown to lead to suppression of [¹⁸F]FDG uptake caused by the inflammatory influx early (i.e. on day 9) after cytotoxic treatment [32]. Thus, the decrease in [¹⁸F]FDG uptake early after treatment that was observed at the same time as the increase in diffusivity (i.e. ADCs), also supports our theory that DWI does indeed capture treatment-induced cell necrosis, rather than inflammatory changes. This is also in agreement with the results of Papaevangelou et al. [31], who observed significant ADC increases in the presence of necrosis in human colon carcinoma xenografts on days 3 and 5 after treatment with irinotecan.

Conversely, the decrease in FDG uptake observed at the ultra-early follow-up in our study was probably also caused, at least in part, by cell death, as indicated by the simultaneous increase in diffusivity and lesion shrinkage. Nevertheless, particularly in the context of HL with its pronounced hyperplastic lymphoid reaction, the anti-inflammatory and lympholytic effects of prednisone on [¹⁸F]FDG uptake and on lesion volume must be taken into account. Furthermore, a previous in vitro study in breast cancer cells established the concept of treatment-induced “stunning” (i.e. a discrepancy between FDG uptake and viable cell number) 24 h after doxorubicin treatment, and thus before the earliest follow-up at 48–72 h after treatment initiation in our study [33]. In the latter study, the stunning effect was explained by declines in GLUT-1 and HKII levels [33]. Similarly, in a murine lymphoma model, a reduction in FDG uptake 16 h after initiation of etoposide treatment was explained by a decrease in the plasma membrane presentation of GLUT-1 and GLUT-3 [23]. It is therefore difficult to judge whether, and to what degree, a possible stunning effect (in combination with the anti-inflammatory effect of prednisone in HL and DLBCL) may have contributed to the observed reduction in FDG uptake in our study, and whether it could partly explain the more pronounced changes in SUV and MTV, compared with the changes in ADC, especially since doxorubicin was also part of the DLBCL and HL treatment regimens in the present study. However, stunning would not have affected ADC and lesion volume changes, which showed the same general trends including differences in response among lymphoma subtypes. Thus, the observed reduction in FDG uptake after treatment initiation can at least not entirely be attributed to stunning and anti-inflammatory treatment effects. Only in three patients was an atypical behaviour observed (initial strong decreases in SUV and MTV from baseline to 48–72 h, followed by increases in SUV and MTV from 48–72 h to 1 week). Unfortunately, in these three patients, we were not able to distinguish between an initial stunning effect that was captured by [¹⁸F]FDG PET at 48–72 h and an inflammatory “flare” reaction that occurred between 48–72 h and 1 week, as this would have required histological evaluation.

Several studies have demonstrated the prognostic value of interim [¹⁸F]-FDG PET, typically performed after two to four therapy cycles, in different lymphoma subtypes, most convincingly in HL [14, 34]. For interim DWI, on the other hand, relatively few data are presently available with regard to its possible value for clinical outcome prediction. Based on a small series of 14 lymphoma patients (12 with DLBCL), De Paepe et al. found that changes in ADC between baseline and follow-up after 2 weeks (as well as 4 weeks) of treatment were significantly correlated with progression-free survival [35]. In a mixed population of 20 patients, 7 with HL and 13 with NHL including aggressive and indolent subtypes, Horger et al. found that changes in ADC after 1 week of treatment predicted outcome as assessed 6 months after the end of treatment [24]. Finally, in a series of 15 patients with FDG-avid MALT lymphoma, Mayerhoefer et al. found that changes in ADC between baseline and interim imaging differed significantly between patients with complete remission and those with residual disease at the end of treatment [10]. To our knowledge, no long-term data regarding the predictive value of therapy-induced ADC changes in lymphoma are presently available.

Our study had several limitations. No posttreatment histological data were available for comparison with the quantitative imaging parameters of the target lesions, because biopsies are rarely performed in routine clinical practice, and purely study-related biopsies were not performed because of ethical considerations. For calculation of MTVs, we used a 41% SUVmax cut-off, as previously recommended [26‐28]. However, it is well-known that major differences in glucose metabolism (and the associated [¹⁸F]FDG uptake) exist between different lymphoma subtypes [36, 37]. We therefore cannot rule out the possibility that the use of a 41% SUVmax threshold led to a small-to-moderate MTV overestimation in patients with indolent lymphomas (i.e. FL, MCL) which frequently show less-pronounced FDG uptake.

In conclusion, our preliminary data clearly suggest that patients with HL and common NHL subtypes may show a substantial response to standard chemo- or immunochemotherapy as early as during the first 48–72 h after treatment initiation, which can be captured in vivo by [¹⁸F]FDG PET/MR. Treatment-induced changes in glucose metabolism appear to precede changes in diffusivity/cell density of lymphoma tissue, which in turn precede changes in lesion volume. Flare reactions that may be detrimental to [¹⁸F]FDG PET appear to be rare events during the investigated time period early after treatment initiation. The degree of treatment response in this very early treatment phase may be histology-dependent (with a trend for stronger responses in more aggressive lymphoma subtypes), but may also depend on the treatment regimen used. Notably, substantial differences exist between patients with regard to this very early treatment response, even between those with the same histological lymphoma subtype and treatment regimen. Consequently, follow-up is warranted to investigate the possible clinical significance of the observed changes in imaging parameters in terms of prognosis and survival.