Aberrant patterns of PET response during treatment for DLBCL patients with MYC gene rearrangements

Purpose MYC gene rearrangements in diffuse large B-cell lymphoma (DLBCL) patients are associated with poor prognosis. Our aim was to compare patterns of 2[18F]fluoro-2-deoxy-D-glucose positron emission tomography computed tomography (PET/CT) response in MYC + and MYC- DLBCL patients. Methods Interim PET/CT (I-PET) and end of treatment PET/CT (EoT-PET) scans of 81 MYC + and 129 MYC- DLBCL patients from 2 HOVON trials were reviewed using the Deauville 5-point scale (DS). DS1-3 was regarded as negative and DS4-5 as positive. Standardized uptake values (SUV) and metabolic tumor volume (MTV) were quantified at baseline, I-PET, and EoT-PET. Negative (NPV) and positive predictive values (PPV) were calculated using 2-year overall survival. Results MYC + DLBCL patients had significantly more positive EoT-PET scans than MYC- patients (32.5 vs 15.7%, p = 0.004). I-PET positivity rates were comparable (28.8 vs 23.8%). In MYC + patients 23.2% of the I-PET negative patients converted to positive at EoT-PET, vs only 2% for the MYC- patients (p = 0.002). Nine (34.6%) MYC + DLBCL showed initially uninvolved localizations at EoT-PET, compared to one (5.3%) MYC- patient. A total of 80.8% of EoT-PET positive MYC + patients showed both increased lesional SUV and MTV compared to I-PET. In MYC- patients, 31.6% showed increased SUV and 42.1% showed increased MTV. NPV of I-PET and EoT-PET was high for both MYC subgroups (81.8–94.1%). PPV was highest at EoT-PET for MYC + patients (61.5%). Conclusion MYC + DLBCL patients demonstrate aberrant PET response patterns compared to MYC- patients with more frequent progression during treatment after I-PET negative assessment and new lesions at sites that were not initially involved. Trial registration number and date of registration HOVON-84: EudraCT: 2006–005,174-42, retrospectively registered 01–08-2008. HOVON-130: EudraCT: 2014–002,654-39, registered 26–01-2015 Supplementary Information The online version contains supplementary material available at 10.1007/s00259-021-05498-7.


Introduction
Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of non-Hodgkin lymphoma with heterogeneous outcome. This heterogeneity can partially be explained by genetic abnormalities, such as MYC oncogene rearrangements, that occur in approximately 10-15% of DLBCL patients and predict outcome independently [1][2][3]. MYC plays a central role in many aspects of the oncogenic process, coordinating metabolism, proliferation, apoptosis, and differentiation of cells. MYC rearrangement is associated with poor outcomes upon standard first-line rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) therapy [1][2][3]. Approximately 30% This article is part of the Topical Collection on Hematology Previously presented at the 33rd Annual Congress of the European Association of Nuclear Medicine -EANM'20 Virtual in the top rated oral presentation (TROP) Session: PET for Haematological Tumours of MYC + large B-cell lymphoma patients have a MYC rearrangement (single hit (SH) DLBCL), whereas in 70% MYC rearrangements are detected together with BCL2 and/or BCL6 rearrangements (double hit (DH) or triple hit (TH) high grade B-cell lymphomas) [4].
In DLBCL, 2[ 18 F]fluoro-2-deoxy-D-glucose positron emission tomography computed tomography ( 18 F-FDG PET/ CT) is the current clinical standard for staging at baseline and treatment evaluation after treatment [5,6]. Both end of treatment PET/CT (EoT-PET) scans and interim PET/CT (I-PET) are reviewed using the Lugano classification criteria (Deauville 5-point scale, DS). To improve consistency of visual assessment, quantitative cut-offs were suggested based on the maximum hepatic standardized uptake value (SUV max ) [6]. In non-selected DLBCL patients, a negative I-PET has a high negative predictive value (NPV) for outcome, but the positive predictive value (PPV) is limited [7,8]. One small study that included 28 MYC + DLBCL patients reported lower NPV values compared to nonselected DLBCL patients [9]. For EoT-PET scans, the NPV is very high in non-selected patients [10]. In MYC + DLBCL patients, the diagnostic performance of I-PET with regard to predicting EoT-PET response patients yielded a NPV of 79%, and a PPV of 60% [11].
Whether molecular high-risk features of DLBCL, such as MYC rearrangement, are correlated with PET responses has not been studied in detail. The aim of this study was to analyze the patterns of PET response in MYC + DLBCL at I-PET and EoT-PET patients and compare it with the response patterns of MYC-DLBCL patients. We also explored the changes in SUV and metabolic tumor volume (MTV) between baseline PET, I-PET, and EoT-PET scans for both MYC + and MYC-DLBCL patients.

Study population
For this analysis, we included 81 MYC + DLBCL patients from the multicenter phase 2 HOVON-130 trial (Eudra-CT: 2014-002,654-39) [11]. All patients included in this trial had confirmation of MYC rearrangement (8q24) by fluorescent in situ hybridization (FISH). BCL2 and BCL6 FISH results were completed when sufficient lymphoma material was available. 18 F-FDG PET/CT scans were performed at baseline, after three cycles of R-CHOP (I-PET) and at end of treatment (EoT-PET). All patients were treated with six cycles of R-CHOP in 21-day cycles combined with lenalidomide, followed by two gifts of rituximab (R). One patient that was included in the HOVON-130 trial had no I-PET or EoT-PET scans available and was therefore excluded from this analysis. DLBCL patients (n = 129) from the multicenter randomized phase 3 HOVON-84 trial (EudraCT: 2006-005,174-42) who had sufficient lymphoma material available for FISH analysis and were MYC-served as a control group [12]. In this trial 18 F-FDG PET/CT scans were performed at baseline, after four cycles of treatment (I-PET) and after 6-8 cycles of treatment (EoT-PET). Patients who were 65 years or younger received eight cycles of R-CHOP intensified with R (RR-CHOP) in 14-day cycles, and those above 65 years received six cycles of RR-CHOP in 14-day cycles. There was no significant difference in survival between both treatment arms, allowing for combined analysis of patients in the HOVON-84 trial.

Qualitative image analysis
For both the HOVON-84 and HOVON-130 trials, 18 F-FDG PET/CT scans were anonymized and uploaded to the Keosys system for web-based viewing and reporting. In total, 20 hospitals included patients for both the HOVON-84 and HOVON-130 trial, and 27 additional hospitals included patients for the HOVON-84 trial. 18 F-FDG PET/CT scans were scanned according to local imaging guidelines. The majority of scans complied with EARL guidelines. All I-PET and EoT-PET scans were centrally reviewed according to the Deauville 5-point scale (DS) [5,6] by two independent, experienced nuclear medicine physicians from the HOVON Imaging Working Group. DS1-3 was regarded as negative (complete metabolic response) and DS4-5 as positive (partial metabolic response or progressive metabolic disease) for both I-PET and EoT-PET. A DS of 5 was assigned when tumor SUV max was ≥ 3 times higher than the hepatic SUV max and/or in case of new lymphoma lesions. Reviewers used an electronic case record form with prespecified nodal (Waldeyer's ring, cervical, supraclavicular, axillary, mediastinum, hilar, paraaortic, mesenteric, spleen, iliac, inguinal, and other) and extranodal localizations (gastrointestinal, central nervous system, skin, liver, lung, pleural, skeletal, and other). Reviewers assigned a DS for individual nodal and extranodal localizations together with a final patient-based DS. A third adjudicator resolved discrepancies between patient-based DS for both studies, and for individual nodal and extranodal localizations for the HOVON-84 study. Reviewers were blinded for clinical outcome.
For the current analysis, discrepancies between DS scores of individual nodal and extranodal localizations for the HOVON-130 study were resolved by an additional adjudication. Moreover, the presence of new lymphomatous lesions between baseline, I-PET, and EoT-PET scans was scored for both the HOVON-84 and HOVON-130 studies. The anatomical localization of new lymphoma lesions were noted according to the predefined nodal and extranodal localizations.

Quantitative image analysis
Quantitative PET/CT analysis was performed using the ACC URA TE tool [13]. Scans were included in this study if they matched the following quality criteria: both PET and low dose CT scans had to be complete, and the liver SUV mean and plasma glucose should be within the ranges suggested by the European Association of Nuclear Medicine guidelines [14]. If the liver SUV mean was outside the suggested ranges, but the total image activity was between 50 and 80% of the total injected activity; these scans were still included.
To assess the changes in uptake and MTV between baseline, I-PET, and EoT-PET for I-PET and EoT-PET-positive lesions, we extracted the MTV and SUV max of these lesions. Baseline MTV of these lesions was calculated using the fixed SUV4.0 segmentation method [15,16]. For PET-positive lesions at I-PET and EoT-PET, MTV was calculated using the method with the most complete tumor segmentation without oversegmentation of non-tumor regions, which could be a fixed threshold (SUV4.0 or SUV2.5), a relative threshold (41%max or 50%peak) or an averaging method (MV2 or MV3). For the MV2 method, voxels detected by ≥ 2 methods (out of SUV4.0, SUV2.5, 41%max, and 50%peak) were selected, and for MV3 method, voxels detected by ≥ 3 methods were selected. For the generated MTV at baseline, I-PET and EoT-PET non-tumor 18 F-FDG avid regions (e.g., bladder, kidney) adjacent to lymphoma lesions were manually removed. All delineations were performed under supervision of an experienced nuclear medicine physician. An increase of 30% or 1.3 units in SUV max , whichever is largest, as cut-off for the smallest detectable change in SUV max . For MTV, we also applied a cut-off of 30% increase [17].

Statistical analysis
The primary endpoint was 2-year overall survival (OS), defined as time from registration to death. Patients still alive were censored at date of last contact.
The Chi-square test for independence was used to compare individual IPI components and patient characteristics, cell of origin, and the differences between PET positive and PET negative proportions at I-PET and EoT-PET between MYC + DLBCL and MYC-DLBCL subgroups. As a sensitivity analysis, MYC + patients were matched with MYC-DLBCL patients based on IPI, Ann Arbor stage, and extranodal involvement. The Chi-square test for independence was used to compare the differences between PET positive and PET negative proportions at I-PET and EoT-PET in this subgroup. Moreover, the Chi-square test for independence was used to compare proportions of patients with new lesions, increased SUV, and increased MTV between MYC + DLBCL and MYC-DLBCL subgroups for EoT-PET compared to I-PET and baseline and I-PET compared to baseline, respectively. A p value of less than 0.05 was considered statistically significant.
Survival curves were obtained with Kaplan-Meier (KM) analyses stratified for PET response at I-PET and EoT-PET for both MYC + and MYC-DLBCL patients. KM curves were compared with log-rank tests. The predictive value of I-PET and EoT-PET was assessed by calculating diagnostic measures (sensitivity, specificity, positive, and negative predictive values (PPV and NPV), respectively) using 2 × 2 contingency tables.
The predictive value of other cut-off values of the DS (DS1-2 vs 3-5 and DS1-4 vs 5) for I-PET and EoT-PET were evaluated in sensitivity analyses by calculation of the sensitivity, specificity, PPV, and NPV.

PET response rates
For the MYC + DLBCL patients, 80 I-PET scans and 80 EoT-PET scans were centrally reviewed. At I-PET, 23 out of 80 patients were PET positive (28.8%, Table 2). Two-year OS for I-PET negative patients was 81.6%, and 60% for I-PET positive patients (p = 0.024; Fig. 1d). At EoT-PET, 26 out of 80 patients were PET positive (32.5%). Thirteen out of 23 I-PET-positive patients remained PET-positive at EoT and 13 out of 56 (23%) I-PET-negative patients converted to PET positivity at EoT. Two-year OS for EoT-PET-negative patients was 92.5%, and 37.8% for EoT-PET-positive patients (p < 0.001, Fig. 1e).
For the MYC-DLBCL patients, 126 I-PET scans and 121 EoT-PET scans were centrally reviewed. At I-PET, 30 out of 126 patients were PET positive (23.8%), which was comparable to the I-PET positivity rate of MYC + DLBCL patients (p = 0.399). Two-year OS for I-PET-negative patients was 91.7%, and 75.4% for I-PET-positive patients (p = 0.016, Fig. 1a). At EoT, 19 out of 121 patients were PET positive (15.7%), which was significantly lower than the EoT PET positivity rate in MYC + DLBCL patients (p = 0.004). Sixteen out of 30 I-PET positive patients remained PET-positive at EoT-PET and 2 out of 91 (2%) I-PET-negative patients with an EoT-PET scan converted to PET-positivity at EoT.
Using the DS1-2 vs 3-5 cut-off resulted in more I-PETand EoT-PET-positive patients in both MYC subgroups (Supplemental Table 1), resulting in lower specificity and PPV and higher sensitivity (Supplemental Table 2). Using the DS1-4 vs 5 cut-off resulted in fewer I-PET-and EoT-PETpositive patients (Supplemental Table 3), leading to higher PPV and specificity and lower sensitivity (Supplemental Table 4).

New localizations
Twenty-one out of 26 EoT-PET-positive MYC + DLBCL patients had new PET-positive localizations at EoT-PET compared to I-PET. Nine EoT-PET-positive patients had new PET-positive localizations that were not initially involved at baseline (1 SH, 6 DH, and 2 TH DLBCL patients) (Fig. 2). Eight out of nine patients had multiple new PET-positive lesions; one patient had a single new PET-positive lesion at EoT-PET compared to baseline ( Table 3). Six of these

Diagnostic performance I-PET and EoT-PET in MYC + and MYC-patients
Using 2-year OS as outcome, the NPV of I-PET and EoT-PET was high (range 81.8-94.1%), and higher for both I-PET and EoT-PET in the MYC-DLBCL patients compared to the MYC + DLBCL patients (Table 4). NPV was highest at EoT-PET for MYC-DLBCL patients (94.1%). PPV was generally low (range 31.6-61.5%), but higher for the MYC + DLBCL patients compared to MYC-DLBCL patients for both I-PET and EoT-PET. PPV was highest at EoT-PET for MYC + patients (61.5%). For the MYC + DLBCL patients, four patients out of 23 patients (17.4%) showed increased lesional uptake, and two patients (8.7%) showed increased lesional MTV at I-PET compared to baseline (Table 5). At EoT-PET, 21 out of 26 patients (80.8%) showed increased lesional MTV and uptake compared to I-PET, whereas 11 out of 26 patients (42.3%) showed increased lesional uptake, and nine (34.6%) patients showed increased lesional MTV at EoT-PET compared to baseline (Table 5).

Discussion
This study shows that MYC + DLBCL patients have a different PET response compared to MYC-DLBCL patients. They experience progressive disease during treatment after I-PET-negative assessment more frequently. Moreover, a substantial part of the MYC + DLBCL patients present with new PET positive lesions at previously uninvolved sites.
In general, I-PET could be a suitable technique to ensure effectiveness of treatment and to exclude the possibility of progression for DLBCL patients [6]. MYC + DLBCL patients have a poor prognosis and relapses are frequent, stressing the need for identification of early markers that reliably predict poor outcomes in this patient population. However, this study showed that the role of I-PET is less evident in this MYC + patient subgroup, as a negative I-PET   Positive Iliac, uterus, 10 Negative Central nervous system assessment results in a positive EoT scan in 22.8% of the patients. We previously reported that in MYC + patients, I-PET response has limited predictive value for EoT-PET response [11]. Nevertheless, in this study, we showed that I-PET can significantly differentiate between responders and non-responders in MYC + DLBCL patients when using 2-year OS as outcome parameter. This is in contrast with the results of 2 other studies, which showed that I-PET did not predict survival in MYC + DLBCL patients [9,18]. Moreover, we showed that I-PET can also significantly differentiate responders and non-responders in MYC-patients. There are no studies that looked at the predictive value of I-PET in MYC-DLBCL patients only. But these results are in line with previous studies, where MYC status was not assessed [7,8,19,20]. In our study, 28.8% of the MYC + patients had positive I-PET scans. This PET-positivity rate was comparable with the PET-positivity rate of MYC-patients (23.8%), and with PET-positivity rates reported in a recent individual patient data analysis [8]. However, Dunleavy and colleagues reported that 70.8% of MYC + DLBCL patients treated with dose-adjusted etoposide, prednisone, vincristine, cyclophosphamide, doxorubicin, and rituximab (EPOCH-R) were PET positive after two cycles of treatment [18]. This positivity rate is much higher than our PET-positivity rates, and the PET-positivity rates of a smaller study incorporating I-PET in MYC + DLBCL patients [9]. This higher PET-positivity rate could not be explained by the DH/TH vs SH ratio, by lower survival rates or by different PET positivity criteria, since all studies used the DS taxonomy. Results of another study that included 51 primary B-cell lymphoma patients treated with dose-adjusted EPOCH-R suggested treatmentrelated inflammation with this treatment-regimen [21]. As the interval between the previous cycle of chemotherapy and I-PET is usually short, this could explain the higher PETpositivity rates in the study of Dunleavy et al. [18]. However, the I-PET-positivity rates in a recent trial, where half of the DLBCL patients were treated with dose-adjusted EPOCH-R [20], did not show higher I-PET positivity rates compared to other studies, contradicting this hypothesis.
Richter et al. showed that MYC + DLBCL patients, with or without BCL2 and/or BCL6 rearrangements, were overrepresented (6/25 vs 21/241) in the I-PET positive group, using the delta SUV response criterion, compared to the I-PET negative group [22]. Furthermore, Yuan et al. also showed that MYC rearrangements were significantly positively correlated with a positive I-PET using delta SUV response criterion, both after 1 and 2 cycles of treatment [23]. These findings were not confirmed in our study: when applying the delta SUV response criterion, 11 MYC + DLBCL patients were positive at I-PET (13.8%), which was not significantly higher than the proportion of I-PET positive patients in MYC-DLBCL patients (9.6%, n = 12).
The EoT-PET positivity rates were significantly higher in the MYC + DLBCL patient group compared to the MYC-DLBCL patient group. As shown by our sensitivity analysis, the significant association between MYC status and PET response remained significant when matching patients on baseline risk, indicating that this higher EoT-PET positivity rate is associated with their MYC status, and not the higher baseline risk of these patients. To the best of our knowledge, no other study reported the EoT-PET positivity rate for solely MYC + DLBCL patients, nor did any study investigate the increase in SUV or MTV between baseline, I-PET, and EoT-PET. Our findings therefore need to be validated in other studies in which both an I-PET and EoT-PET for MYC + DLBCL patients will be performed.
Our study shows that the NPV was high at I-PET and EoT-PET for MYC-DLBCL patients. This NPV is higher than the NPV of I-PET of most DLBCL studies that included both MYC-and MYC + DLBCL patients [7], indicating that further stratification of patients using MYC status improves the NPV of I-PET and EoT-PET scans for DLBCL patients. The PPV of both I-PET and EoT-PET is lower, especially for MYC-patients. For MYC + patients, the PPV at EoT-PET is relatively high.  I-PET  2  21  0  30  I-PET vs EoT-PET  21  5  8  11  Baseline vs EoT-PET  9  17  1  18  SUV  Baseline vs I-PET  4  19  1  29  I-PET vs EoT-PET  21  5  6  13  Baseline vs EoT-PET  11  15  2  17 To the best of our knowledge, this is the first study to make a head-to-head comparison of PET response patterns of MYC + DLBCL and MYC-DLBCL patients. One of the strengths of this study is the fact that the PET scans and pathology data were centrally reviewed. The uniform interpretation of scans and pathology in both studies resulted in high-quality data. Moreover, with a total of 214 patients included, of which 82 had a MYC translocation, the sample size of this study is fairly large allowing firm conclusions. We choose OS as outcome parameter, instead of the more commonly used progression free survival (PFS) because progression for the MYC + DLBCL patients was often confirmed using the EoT-PET results. Therefore, almost all MYC + DLBCL patients with a positive EoT-PET had a PFS event at EoT-PET. Furthermore, SH and DH/TH DLBCL patients were included in our MYC + DLBCL group. Unfortunately, our subgroups of SH and DH/TH patients were too small to draw conclusions per subgroup. This study has some limitations. First, MYC + patients received intensified chemotherapy compared to the MYC-patients. This treatment of MYC + patients is believed to be more effective than R-CHOP without intensification. Therefore, if these patients would receive R-CHOP without intensification, the patterns of PET response between MYC + and MYC-groups could be different. Secondly, the timing of I-PET differed between subgroups. A recent individual patient data analysis showed that I-PET positivity using DS criteria at I-PET3 and I-PET4 was almost identical (21.7 vs 21.2%) [8], implying that the I-PET-positivity rate for MYC-DLBCL patients would be similar after three cycles of treatment. However, a small effect of timing cannot be precluded in our results. Thirdly, new PET + localizations at EoT were not biopsy proven, but at least one new localization in each patient was located at sites that are not suspicious for infections or other causes. Lastly, not all patients in the HOVON-84 study had enough biopsy material to confirm their MYC status, which could have led to selection bias. However, when matching MYC + and MYC-patients on baseline characteristics, we still found similar results, implying that selection bias did not influence our results to a large extent.
Future studies should focus on the potential of other quantitative PET measures or molecular biomarkers to stratify "good" and "poor" responders within MYC + DLBCL patients. Robust and easy to use biomarkers for early identification of poor responders in this patient group are essential and an integrative approach with both molecular data and quantitative PET metrics could improve prediction of prognosis and guide choice of therapies.

Conclusion
MYC + DLBCL patients demonstrate different response patterns compared to MYC-DLBCL patients with more progressive disease during treatment after I-PET negative assessment, with more nodal and extranodal lesions at sites that were not initially involved.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.