Correlation Between PET/CT Parameters and
KRAS Expression in
Colorectal Cancer
Shang-Wen Chen, MD,*Þþ Hua-Che Chiang, MD,§ William Tzu-Liang
Chen, MD,Þ§
Te-Chun Hsieh, MD,||? Kuo-Yang Yen, MD,||? Shu-Fen Chiang, MD,**
and Chia-Hung Kao, MDÞ||
C
olorectal cancer (CRC) is a frequently occurring malignancy.1Over the last 2 decades, multimodality treatment has resulted in crucial improvements for the treatment of this disease. Recently, new molecular insights and technologies have indicated the initiation, early detection, and prognostic markers of and effective drug therapy for CRC. KRAS mutations, which occur in approximately 40% of
CRCs, are particularly crucial because they can predict a lack of responses to therapies with antibodies targeted to the epidermal growth
factor receptor (EGFR).2Y5 18F-FDG PET imaging is widely used for
diagnosis, monitoring treatment response, surveillance, and prognostication for CRC. Despite imaging techniques being critical in the
preoperative workup for treatment decisions, a paucity of correlation studies exists between pretreatment image findings and genomic expression in this patient setting.
A previous study indicated that SUVmax for the primary tumor and the tumor-to-liver ratiowas higher in CRC with KRAS mutations.6
To optimize the predictive values, however, 2 questions remain unanswered. First, the individual predictive performance across various
approaches of PET/CT-related parameters, such as metabolic tumor volume (MTV) or total lesion glycolysis (TLG), remains unknown. Second, CRC is considered a heterogeneous, complex disease that comprises various tumor phenotypes,7 and accumulating evidence
suggests that grouping these anatomically distinct diseases could be a clinical and biological oversimplification.8 Thus, whether the predictive
clarification. Because of a research gap in matching comprehensive PET/CT parameters and KRAS expression in CRC patients, this study compares various autosegmentation methods with KRAS mutations, to determine the most effective approach to differentiate mutant and wild-type CRC. The results supplement genomic analysis to determine the optimal therapeutic strategies for CRC patients by predicting tumor response to various treatment modalities.
MATERIALS AND METHODS
Patient Population
The 121 newly diagnosed CRC patients scheduled to undergo curative surgical procedures at China Medical University Hospital between January 2009 and December 2012 were included in this retrospective study (certificate number of local institutional review board: DMR99-IRB-010-1). Tumor locations were colon or sigmoid colon (72 patients) and rectum or rectosigmoid junction (49 patients). The median age was 59 years (range, 26Y86 years). Seventy-two patients were men, and 49 were women. All patients received PET/ CT for pretreatment staging and underwent primary tumor resection thereafter (median, 7 days; range, 1Y26 days). No patient received preoperative chemotherapy or had a history of diabetes. All patients had a normal serum glucose level prior to obtaining PET/CT images. The characteristics of the 121 patients are shown in Table 1.
PET/CT Image Acquisition
All patients fasted for at least 4 hours prior to 18F-FDG PET/
CT imaging. The images were captured using a PET/CT scanner (PET/CT-16 slice, Discovery STE; GE Medical System, Milwaukee, Wis) approximately 60 minutes after administering 370 MBq of 18FFDG.
Patients were requested to rest during the uptake period. The FDG-PET data were inputted into the workstation, and the images
were reviewed to localize the target lesions, as confirmed by 2 nuclear medicine physicians. The physicians were unaware of the information
of the preoperative images. The PET/CTworkstation provided a quantification of FDG uptake for SUV. Nuclear medicine physicians identified the locations of SUVmax and the values for the primary tumors. This procedure was detailed in our previous report.9
Measurement of MTV and TLG
The PET-based MTVs were measured from attenuationcorrected FDG-PET images, using an SUV-based automated contouring
program (Advantage Workstation Volume Share version 2; GE Health, Milwaukee, Wis). The MTVs were measured from attenuation-corrected FDG-PET images. The MTV was defined as the sum of metabolic volumes of the primary tumors. The volume boundaries were drawn sufficiently widely to incorporate each target lesion in the axial, coronal, and sagittal FDG-PET images. To define the contouring margins around the tumor, we used SUVmax of 2.5 (MTV2.5) and SUVmax of 3.0 (MTV3.0), as previously reported.10
Furthermore, volumes greater than 20% of SUVmax (MTV20%), 30% of SUVmax (MTV30%), 40% of SUVmax (MTV40%), and 50% of SUVmax (MTV50%) were analyzed. The contour around the target lesions within the boundaries was automatically produced, and the voxels presenting an SUV intensity of SUVmax Q2.5 or Q3.0, and Q20%, Q30%, Q40%, or Q50% of SUVmax within the contouring margin were incorporated to define the tumor volumes. The MTVs for the primary tumors included adjacent lymph nodes with small volumes. The small lymph nodes adjacent to the primary tumor cannot be segmented from the primary tumor by PET/CT when they appear similar to a part of the primary lesion. However, large lymph nodes neighboring the primary tumor can be segmented using an automatic volume-of-interest tool on PET/CT, even if they are partially contiguous to the primary tumor.
The TLGs were also measured from attenuation-corrected FDGPET images, using an SUV-based automated contouring program (Advantage Workstation Volume Share version 2; GE Health). The TLG was calculated according to the following formula: TLG = mean SUV _ MTV.11 We used the same threshold levels as the MTVs,
namely, TLG2.5, TLG3.0, TLG20%, TLG30%, TLG40%, and TLG50%.
Measurement of PET-Based Maximal Tumor Width
The PET-based maximal tumor width (TW) was also measured from attenuation-corrected FDG-PET images, using an SUV-based automated contouring program. To define the maximal TW around the tumor, we used the measured distance greater than 30% of SUVmax (TW30%), 40% of SUVmax (TW40%), and 50% of SUVmax (TW50%). Using this tool, the calculated unit for the analyses was millimeters. The details were described previously.12
KRAS Mutation Analysis
resection, and the tissue blocks were reviewed by the pathologists to select the tumor area. DNA was extracted from 5-mm
formalin-fixed, paraffin-embedded tumor tissue slides, using the QuickExtract FFPE DNA Extraction Kit (Epicentre Biotechnologies, Madison, Wis). KRAS exon 2 was amplified using polymerase chain reaction and was analyzed using direct sequencing with the ABI 3730XL automated DNA analyzer (Applied Biosystems, Foster City, Calif ) according to the manufacturer’s instruction.
Statistical Analysis
All values are expressed as means T SD. Differences in
SUVmax or various thresholds of PET/CT-related parameters between mutated and wild-type KRAS were tested using a
Mann-Whitney U test. Categorical variables between the 2 groups were
assessed using a W2 test. The analyses were tested by receiver operating
characteristic curve analysis, to compare the predictive ability. In addition, the predictive values of PET-related parameters for the KRAS status were examined using multivariate logistic regression analysis. All analyses were 2-sided, with P G 0.05 considered statistically significant. Statistical analyses were performed using SPSS,
version 13.0 (SPSS Inc, Chicago, Ill).
RESULTS
Correlation Between KRAS Expression and
PET/CT-Related Parameters
The patient numbers for KRAS mutant and wild type were 49
and 72, respectively. The details of the PET/CT-related parameters are summarized in Supplemental Table, http://links.lww.com/CNM/A6. Except for the methods using MTV20%-50% and TW50%, certain threshold methods showed a significant trend for CRC tumors with mutated KRAS to have higher accumulation of FDG uptake compared with wild type. The results are listed in Table 2. An example with a dense accumulation of FDG within the tumor is illustrated in Figure 1.
Predictive Value of PET/CT-Related Parameters for
the KRAS Mutant
Receiver operating characteristic curves were analyzed to compare the efficacy of various methods for determining thresholds for autosegmentation contouring, and the results showed that SUVmax, MTV3.0, TW40%, and TLG30% predicted the KRAS mutant most
curve were 0.65 T 0.05, 0.64 T 0.05, 0.64 T 0.05, and
0.64 T 0.05. Thus, they were chosen as variables for the multivariate analysis. As summarized in Table 3, logistic regression analysis showed that SUVmax and TW40% were the 2 predictors of KRAS mutations. The odds ratio (OR) was 1.23 for SUVmax (P = 0.02; 95% confidence interval [CI], 1.01Y1.52) and 1.15 for TW40% (P = 0.02; 95% CI, 1.02Y1.30). Figure 2 depicts the quantitative difference of SUVmax and TW40% between the 2 groups. To clarify the correlation between the 2 predictors, a linear correlation test showed no apparent
relation (R2 = 0.009). In addition, whereas the pretreatment carcinoembryonic
antigen level showed a marginal impact, no association
existed between pathological T or N staging and KRAS expression. We then sought to determine the optimal cutoff to distinguish between the 2 groups. Receiver operating characteristic analysis showed the highest accuracy (70%) with an SUVmax cutoff value of 11. The sensitivity and specificity for predicting the KRAS mutant were 52.4% and 71.7%, respectively (positive predictive value = 65.3%, negative predictive value = 59.7%). Using the median value of TW40% as a cutoff (2.6 cm), the sensitivity, specificity, and accuracy were 53.2%, 67.6%, and 62%, respectively.
Difference in FDG Accumulation Between Colon and
Rectum in Predicting KRAS Mutant
Based on the location of primary tumors, patients were divided
into 2 groups: patients with cancer in the colon or sigmoid colon (n = 72) and those with cancer in the rectum or rectosigmoid junction (n = 49). Using the Mann-Whitney U test, the SUVmax value remained statistically significant in predicting KRAS mutations in the former (P =
0.005). However, TW40% did not reflect the genetic mutant for this group (P = 0.06). When using the optimal cutoff value of SUVmax at 11, the sensitivity and specificity for predicting the KRAS mutant were 54.3% and 81.0%, respectively (positive predictive value = 73.1%, negative predictive value = 65.2%, accuracy = 68.1%). In patients with rectum or rectosigmoid junction cancer, TW40% was significantly higher in the mutant group (P = 0.011). When using the median value of TW40% (2.4 cm) as a cutoff, the sensitivity, specificity, and accuracy were 80%, 79.1%, and 71.4%, respectively. In contrast,
SUVmax failed to differentiate between the 2 groups (P = 0.54).
Oncogenic activation of KRAS can influence several cellular
processes that regulate biological course,10 and KRAS mutations occur
in several human malignancies, including pancreatic cancer, nonYsmall cell lung cancer, and CRC. From a clinical perspective, predictors of treatment outcomes for CRC patients who are candidates for anti-EGFR monoclonal antibody therapies have become a therapeutic standard. Two anti-EGFR monoclonal antibodies (cetuximab and panitumumab) have been suggested for the treatment of metastatic CRC for tumors without KRAS mutations. Two studies have shown that anti-EGFR monoclonal antibodies have significant efficacy in the
treatment of metastatic CRC patients with wild-type KRAS tumors.13,14 In contrast, the
American Society of Clinical Oncology suggested that
patients with metastatic CRC, having a KRAS mutation in codon 12 or 13, should not receive anti-EGFR antibody treatment.15
Although interest is developing in the role of FDG-PET in
staging or monitoring response in CRC, FDG-PET has rarely been investigated in genomic expression. A unique advantage of FDG-PET scanning is the ability to use the quantitative information of the glucose uptake within the tumor or to automatically create a contour
around the tumor. In contrast with CT or magnetic resonance imaging, this autocontouring process substantially reduces the interobserver variability in the interpretation of images.16Y19 Mechanisms affecting
FDG accumulation in cancer tissues are complex.20,21 In CRC, certain
studies have suggested that glucose transporter 1Ymediated FDG accumulation is more essential than hexokinase type II activity.6,22 In this
study, we report that SUVmax and FDG accumulation of several
thresholds were higher in mutated KRAS, by examining the comprehensive approaches of PET/CT-related parameters. In addition,
we highlight the geographical differences in the predictive performance between SUVmax and TW40%. Kawada et al6 conducted a pilot predictive
study by using SUVmax and tumor-to-liver ratio in a cohort of 51 CRC patients. They showed that SUVmax had an OR of 1.17 with an accuracy of 75% in forecasting mutated KRAS when using a cutoff value of 13. Using a large sample size and various threshold methods, we demonstrated the differences across these PET/CT-related parameters in predicting genomic expression. Particularly, theTW40%method
can achieve higher accuracy when applied to predicting rectal cancer. Most studies that have been proposed to predict prognosis in
CRC patients have combined colon and rectal cancers, but whether this combination is appropriate is unknown. Differences in the expression of specific genes have been reported; for example, colon
cancers have a higher number of mutations, including KRAS and
BRAF.23 In our data, no obvious difference existed between the percentage
of mutated KRAS from the 2 sites (colon: 36.1%; rectum:
46.9%). In addition, we observed no quantitative variation of the PET/CT-related parameters between the two. The lack of a significant association between the SUVmax and the TW40% suggests that TW40% could be considered a novel approach for predicting KRAS mutations. In particular, this value can be more accurate when applied to predicting tumors of the rectum. Theoretically, TWs may
reflect macroscopic tumor burden more precisely than a single point, because the TWs represent a range of maximal TW by using a fixed threshold. Further studies are essential to validate our findings. This research should be interpreted with 2 considerations.
First, CRC tumors with mutated KRAS showed only 1.23-fold increases in SUVmax with an accuracy of 70%. Combined with the
results from Kawada and colleagues’ study,6 currently, FDG-PET/CT
is not sufficient for replacing the mutational testing. Second, heterogeneity of KRAS status within a primary CRC tumor has been
reported.24 As a result, the correlation study might be biased because
dissected specimens for mutational testing may not reflect the exact macroscopic status of the entire tumor, and PET/CT may represent the gross status of the tumors.6 To optimize the therapeutic effect of
PET/CT in CRC, future studies should include more participants prospectively and use standardized protocols for FDG-PET acquisition and correction of the partial volume effect or false-positive PET findings.25,26 In addition, together with more comprehensive genomic
information such as BRAF or HMGA2,27 it is imperative to understand whether FDG-PET/CT
scans might predict the actual response to
anti-EGFR regimens, in addition to survival rates. Furthermore, given several PET/CT-relative parameters were shown to be correlated with
pathological response after neoadjuvant chemoradiotherapy in patients with locally advanced rectal cancer,28,29 there is a need to investigate
the association between genomic expression and quantitative change of FDG-PET for the responders and the nonresponders.
PET/CT-related parameters can be used for supplementing
genomic analysis to determine KRAS expression in CRC. The mutated KRAS tumors are associated with higher FDG accumulation
across several threshold methods. SUVmax and TW40% are the 2 predictors of KRAS mutations. The accuracy of SUVmax was superior in patients with colon or sigmoid colon cancers, whereas
