Relationship Between Tumor Mutational Burden and Maximum Standardized Uptake Value in PET (Positron Emission Tomography) Scan in Cancer Patients

Purpose: Deriving links between imaging and genomic markers is an evolving eld. 18 F-FDG PET/CT ( 18 F-uorodeoxyglucose Positron Emission Tomography- Computed Tomography) is commonly used for cancer imaging, with maximum standardized uptake value (SUV max ) as the main quantitative parameter. Tumor mutational burden (TMB), the quantitative variable obtained using next-generation sequencing on a tissue biopsy sample, is a putative immunotherapy response predictor. We report the relationship between TMB and SUV max , linking these two important parameters. Methods: In this pilot study, we analyzed 1923 patients with diverse cancers and available TMB values. Overall, 273 patients met our eligibility criteria in that they had no systemic treatment prior to imaging/biopsy, and also had 18 F-FDG PET/CT within six months prior to the tissue biopsy, to ensure acceptable temporal correlation between imaging and genomic evaluation. Results: We found a linear correlation between TMB and SUV max (p<0.001). In the multivariate analysis, only TMB independently correlated with SUV max whereas age, gender and tumor histology did not. Conclusion: Our observations link SUV max in readily available, routinely used, and non-invasive 18 F-FDG PET/CT imaging to the TMB, which requires a tissue biopsy and time to process. Since higher TMB has been implicated as a prognostic biomarker for better outcomes after immunotherapy, further investigation will be needed to determine if SUV max can stratify patient response to immunotherapy.


Introduction
Creating a link between imaging ndings and genomic data in patients with cancer is crucial in the evolving world of genomics. Radiologic markers have shown promise for non-invasive identi cation of molecular properties [1]. Imaging markers can provide surrogate genomic markers from imaging data for diagnosis, prognosis and strati cation of cancer patients in the emerging eld of personalized medicine.
Such links between imaging and genomics have been formed in computed tomography (CT) and magnetic resonance imaging (MRI), whereas fewer studies have investigated such relationship in 18 Fuorodeoxyglucose ( 18 F-FDG) positron emission tomography (PET) [2][3][4]. 18 F-FDG PET/CT is standard of care and plays a pivotal role in cancer diagnosis and staging [5]. The maximum standardized uptake value (SUV max ), a relative measure of FDG uptake, is the most widely used quantitative parameter for the assessment of cancer patients [6,7]. Traditionally the SUV max has been correlated with histopathological ndings. As the era of personalized medicine continues to move rapidly toward molecular strati cation, there have been studies to associate SUV max with biologic pathways, although SUV max is still generally unspeci ed at a molecular level [4,8].
Tumor mutational burden (TMB) is de ned as the total number of somatic mutations identi ed per megabase pair (Mbp) of coding area and is quanti ed using next-generation sequencing (NGS). Tumors with high TMB likely harbor numerous neoantigens created from the mutanome, possibly eliciting an endogenous immune response or enhancing tumor metabolic rate, which may explain the correlation of high TMB and response to checkpoint blockade immunotherapy [9][10][11][12].
Herein we demonstrate a positive relationship between SUV max and TMB in patients with cancer. Our nding links SUV max in readily available, routinely used, and non-invasive 18 F-FDG PET/CT imaging to the TMB, which requires a tissue biopsy and time to process. Since higher TMB has been suggested to predict better outcomes after immunotherapy, further prospective study of this association and its implications for cancer immunotherapy are needed.

Patient Characteristics:
Of 1923 patients in the database, we found 273 patients with metastatic cancer had no systemic treatment prior to imaging/biopsy, and also had 18 F-FDG PET/CT within 6 months prior to the tissue biopsy (Table 1). Based on the TMB values and prior precedent for cut-off values [13], patients were categorized into three groups: TMB = 0-1 mutations/mb (N = 39 patients, 14%), TMB = 2-11 mutations/mbp (N = 174, 64%), and TMB ≥ 12 mutations/mb (N = 60, 22%); the mean TMB in each group was 0, 5, and 37 mutations/mbp. There was no difference in gender distribution between groups. Patients in the TMB = 2-11 mutations/mb group were slightly younger on average than those in the two other TMB groups (63.9 vs. 68.5 or 68.3 years, p = 0.03). Histology distribution showed statistically signi cant differences between TMB categories in different cancer types, with higher TMB in melanoma and lung cancer than in breast, gastrointestinal or "other" cancers ( Table 1). The most notable pattern was seen with melanoma, which expectedly had the highest percentage of patients among cancer types in the TMB ≥12 mutations/mb category (60% vs. 7%-30% in other cancer types) [12,14].  *Other cancers consisted of head and neck, adrenal, bladder, ovary, uterus, prostate, musculoskeletal, and hematologic malignancies, and cancers of unknown primary.
SUV max correlates with TMB: Median SUV max was 3.9, 6.8 and 9.2 for the 0-1, 2-11 and >12 mutations/mb groups (p < 0.0001). Raw diagnostics showed that the TMB is the only variable with statistically signi cant relationship with SUV max (p<0.003) ( Table 2). However, due to the highly skewed distributions of both TMB and SUV, shifted log transformations were used for analysis because statistical model diagnostics indicated that both SUV max and TMB should be analyzed on the log scale (  (Figure 1). Linear correlation between all shifted-log TMB and shifted-log SUV max had Pearson correlation coe cient r=0.34, (p<0.001) ( Figure 2). Among different cancer types, breast cancer patients showed linear correlation between shifted-log TMB and shifted-log SUV max with Pearson correlation coe cient r=0.40, (p=0.008). This linear correlation coe cient for the lung cancer patients was r=0.43, (p=0.001), and for the other cancer patients was r= 0.37 (p<0.001). For the melanoma and gastrointestinal patients this relationship was not statistically signi cant, perhaps due to smaller number of patients in these two groups (N= 15, and 36, respectively).  ***Only variables with p-value ≤ 0.1 in univariate were tested in multivariate analysis.

Discussion
To our knowledge, this is the rst study to investigate the relationship between SUV max and TMB in patients with diverse cancers. Our hypothesis was that higher mutational load (as re ected by TMB) might correlate with metabolic recon guration [15], and immune in ammatory response [12], and that either of these features would be associated with a higher SUV max [16]. Our study con rmed that higher TMB was the only evaluated variable that independently correlated with higher SUV max . Of interest, one prior study examined this question, albeit in lung cancer alone [4]. They found no signi cant relationship between SUV max and TMB. However, there were some major differences between their study and ours: (i) Moon and colleagues con ned their observations to lung cancer, whereas our study included a variety of malignancies; and (ii) they did not note the timing of the FDG PET-CT versus the biopsy [4]. In our study, the biopsy was taken ≤ 6 months before the PET scan. Various cutoffs have been previously established for TMB, including the dichotomization at 12 mutations/mb, to be predictive of immunotherapy response [13]. Upon categorizing patients into three groups based on TMB levels: 0-1, 2-11 and ≥ 12 mutations/mb groups, we found that patients in the higher TMB group have higher SUV max values and this difference was statistically signi cant between all three groups with p < 0.0001 (Table 1).
To con rm that the relationship between TMB and SUV max is independent of confounders, we analyzed the data in multivariate models. The only parameter that showed a signi cant relationship with SUV max was TMB (multivariate p < 0.0001), con rming the independent correlation between TMB and SUV max (Table 3). Further, there was a linear relationship between TMB and SUV max in the log-scale (r = 0.34, p < 0.001) (Fig. 2). Sex, age and cancer type had no statistically signi cant association with SUV max (Table 3).
Several genomic alterations have been related to immunotherapy response, including but not limited to microsatellite instability high (MSI-H) status (which results in high TMB), high TMB itself, PBRM1 mutations and APOBEC-related mutagenesis [12,13,[17][18][19]. TMB varies dramatically between tumor types, with skin and lung cancers, having higher median TMBs than most other cancers [20,21]. Our previous studies indicated that the median TMB for responders vs. non-responders to anti-PD-1/PD-L1 monotherapy was 18.0 vs. 5.0 mutations/Mb, with higher TMB predicting favorable outcomes across diverse tumors [12,22]. Other studies have found that higher TMB was linked to improved survival following immunotherapy in diverse cancers for the top 20% of TMBs in each histology [23]. Various investigations have used different cut offs for de ning the relationship between TMB and checkpoint blockade response [24] and our own work has suggested a linear correlation between TMB and response [12].
We hypothesize that higher TMB promotes metabolic recon guration, causing increased glucose metabolism rate (GMR), and thus higher SUV max . Carbohydrate metabolism has been previously shown to have correlation with TMB [15]. GMR-TMB correlation could explain our nding of SUV max -TMB correlation, although the exact mechanism for this nding is not understood [15]. An alternative explanation for the correlation between TMB and SUV max might be based on an innate immune response to tumors with higher TMB. Indeed, higher TMB correlates with better response to immune checkpoint blockade and it is conceivable that innate immunity might also be triggered in the presence of high mutational load. An immune cell in ltrate would create increased glycolytic activity and an in ammatory response that would manifest as higher SUV max [25]. We have also previously shown increased SUV max in tumors with higher number of characterized genomic alterations [26] consistent with this work.
Our study had several important limitations: rst it is a retrospective analysis and thus TMB and SUV max parameters were not fully synchronized; second, although the full cohort included 1923 patients, only 273 patients had PET scans within six months before their biopsies for TMB; third, we do not know the mechanism underlying the relationship between TMB and SUV max; fourth this study was singlecenter/single-camera, and fth, a variety of tumor types were included in the analysis, though the latter two may also suggest the homogeneity of PET results and generalizability of results across histologies, respectively); and fourth, we did not examine variant genes or molecules associated with SUV max , which could be key markers of SUV max [27]. Future studies are needed to expand the number of the patients, and to evaluate such relationship in each individual cancer type. Multi-center study and also same-day PET scans/biopsy for TMB are needed would be needed to validate our ndings. Furthermore, exploring the direct relationship between SUV max and immunotherapy response is future step since we found SUV max is correlated with TMB and it is known that higher TMB is correlated with immunotherapy response.

Patient selection:
We performed a search and found 1923 patients who had TMB values on biopsy tissue samples obtained by hybrid capture-based NGS (Foundation Medicine) at UC San Diego Moores Cancer Center.
Among those, 273 patients had no systemic treatment prior to imaging/biopsy, and also had 18 F-FDG PET/CT within 6 months prior to the tissue biopsy, to ensure acceptable temporal correlation between imaging and genomic evaluation.
18 F-FDG PET-CT imaging: All patients received PET imaging under standard conditions as needed for their disease assessment.
Patients were asked to fast for at least six hours prior to their scan. Blood glucose levels were measured immediately before the FDG injection and no patient had a blood glucose level >160 mg/dl. Patients were injected with 370 -740 MBq FDG, intravenously, within 5-10 seconds. Following an uptake period of approximately one hour in a quiet room at rest, multi-station 3-dimensional (3D) whole body PET acquisition with CT (for attenuation correction) was performed for approximately 60 min, using the same GE Discovery VCT scanner (GE, Waukesha, WI) for all the patients. The scanner was in compliance with American College of Radiology guidelines. Whole-body CT covered the region from the head to the midthigh. PET images were acquired, after the CT scan, at a rate of 2 minutes/bed position, in the 3 dimensional (3D) acquisition mode. CT images were then reconstructed onto a 512 x 512 matrix. PET images were reconstructed using a standard whole body 3D iterative reconstruction: 2 iterations; 28 subsets onto a 128 x 128 matrix with attenuation correction, decay correction, and scatter correction. The photon energy window was 425 -650 keV. Slice thickness was 3.27 mm and reconstruction diameter was 70 cm. Pixel size was 5.47 mm x 5.47 mm with spatial resolution of 5 mm. 18 F-FDG PET/ CT images were generated for review on a workstation.
Image Analysis: All PET images were interpreted on the institution's pictures archiving and communication system (PACS), (AGFA Impax 6.3, Mortsel Belgium) by a board-certi ed academic nuclear medicine physician/radiologist and veri ed by a second nuclear medicine physician/radiologist. Focal activities of the lesions were manually identi ed on the PET images. SUVs of the lesions were obtained by manually placing a circular region of interest (ROI) at the site of the maximum FDG uptake in the PET images and the maximal activity (SUV max ) was recorded. The SUV was calculated as decay-corrected activity of tissue volume (kBq /mL)/injected FDG activity per body mass (kBq/g). In most of the cases, the biopsied lesion was selected for analysis; however if the biopsied lesion was smaller than 1 cm, the most FDG-avid lesion, larger than 1 cm, was selected, to avoid partial volume effect. Therefore all the lesions that underwent SUV max analysis, were > 1 cm diameter. For patients showing no focal 18 F-FDG uptake on PET, a rounded SUV max of 0 was recorded. It should be noted that those patients with only background uptake have no elevated glucose uptake; the exact SUV max number may vary in these different patients due to technique and background, so they were all rounded to 0, for a more accurate representation.

Evaluation of TMB:
Formalin-xed, para n-embedded tumors were submitted for NGS to Foundation Medicine (clinical laboratory improvement amendments (CLIA)-certi ed lab). The Foundation One assay was used (hybridcapture-based NGS; 182, 236 or 315 genes, depending on the time period). The methods have been previously described [28]. Average sequencing depth of coverage was > 250×, with >100× at >99% of exons. For TMB, the number of somatic mutations detected by interrogating 1. years and was also used as the response variable in ANOVAs with sex and cancer type. The variables with p<0.1 in these four analyses were then used in a general linear model with SUV max as the response variable. Differences between groups were considered to be signi cant at a p-value < 0.05 and con dence intervals (CI) were done at con dence level 95%. The geometric mean was used for some analysis (geometric mean of N numbers is the nth root of the product of the numbers). Data are reported as mean standard deviation (SD).

Conclusion
we found a linear positive correlation between TMB and SUV max in diverse cancers. Of the features evaluated, multivariate analysis showed TMB to be the only factor independently associated with SUV max . Future prospective studies with PET scans and biopsy for TMB done on the same day are needed to validate the ndings in this area. Furthermore, it will be important to determine if tumors with higher SUV max respond better to immunotherapy, as might be expected, since higher TMB is correlated with immunotherapy response [12].  percentiles. The circles represent outlier SUVmax values, de ned as either larger than Q3 + 1.5 × IQR or smaller than Q1 -1.5 × IQR, where IQR = Q3 -Q1 is the interquartile range. The horizontal "whiskers" represent the largest and smallest non-outlier observations in the data set. All p-values are from analysis on log scale.