Eur J Nucl Med Mol Imaging (2013) 40:853–864 DOI 10.1007/s00259-013-2357-3

ORIGINAL ARTICLE

Fifteen different 18F-FDG PET/CT qualitative and quantitative parameters investigated as pathological response predictors of locally advanced rectal cancer treated by neoadjuvant chemoradiation therapy Anna Margherita Maffione & Alice Ferretti & Gaia Grassetto & Elena Bellan & Carlo Capirci & Sotirios Chondrogiannis & Marcello Gava & Maria Cristina Marzola & Lucia Rampin & Claudia Bondesan & Patrick M. Colletti & Domenico Rubello

Received: 23 November 2012 / Accepted: 24 January 2013 / Published online: 16 February 2013 # Springer-Verlag Berlin Heidelberg 2013

Abstract Purpose The aim of this study was to correlate qualitative visual response and various PET quantification factors with the tumour regression grade (TRG) classification of pathological response to neoadjuvant chemoradiotherapy (CRT) proposed by Mandard. Methods Included in this retrospective study were 69 consecutive patients with locally advanced rectal cancer (LARC). FDG PET/CT scans were performed at staging and after CRT (mean 6.7 weeks). Tumour SUVmax and its Anna Margherita Maffione and Alice Ferretti contributed equally to the preparation of the manuscript. A. M. Maffione : A. Ferretti : G. Grassetto : S. Chondrogiannis : M. C. Marzola : L. Rampin : C. Bondesan : D. Rubello Nuclear Medicine Department, PET unit, Santa Maria della Misericordia Hospital, Rovigo, Italy A. Ferretti : E. Bellan : M. Gava Medical Physics Department, Santa Maria della Misericordia Hospital, Rovigo, Italy C. Capirci Radiotherapy Department, Santa Maria della Misericordia Hospital, Rovigo, Italy P. M. Colletti Department of Radiology, University of Southern California, Los Angeles, CA 90033, USA A. M. Maffione (*) SOC Medicina Nucleare, Santa Maria della Misericordia Hospital, Viale Tre Martiri, 140, 45100 Rovigo, Italy e-mail: [email protected]

related arithmetic and percentage decrease (response index, RI) were calculated. Qualitative analysis was performed by visual response assessment (VRA), PERCIST 1.0 and response cut-off classification based on a new definition of residual disease. Metabolic tumour volume (MTV) was calculated using a 40 % SUVmax threshold, and the total lesion glycolysis (TLG) both before and after CRT and their arithmetic and percentage change were also calculated. We split the patients into responders (TRG 1 or 2) and nonresponders (TRG 3–5). Results SUVmax MTV and TLG after CRT, RI, ΔMTV% and ΔTLG% parameters were significantly correlated with pathological treatment response (p<0.01) with a ROC curve cut-off values of 5.1, 2.1 cm3, 23.4 cm3, 61.8 %, 81.4 % and 94.2 %, respectively. SUVmax after CRT had the highest ROC AUC (0.846), with a sensitivity of 86 % and a specificity of 80 %. VRA and response cut-off classification were also significantly predictive of TRG response (VRA with the best accuracy: sensitivity 86 % and specificity 55 %). In contrast, assessment using PERCIST was not significantly correlated with TRG. Conclusion FDG PET/CT can accurately stratify patients with LARC preoperatively, independently of the method chosen to interpret the images. Among many PET parameters, some of which are not immediately obtainable, the most commonly used in clinical practice (SUVmax after CRT and VRA) showed the best accuracy in predicting TRG. Keywords PET/CT . Locally advanced rectal cancer . Neoadjuvant chemoradiotherapy . TRG . SUV . TLG

854

Introduction Colorectal carcinoma is the third most common cause of malignancy in both men and women and the second most common cause of death in the western world, with an increased incidence due to the diffusion of screening programmes and procedures [1–3]. The most important treatment approach is surgery which in the early stage (stage I) can be curative. For locally advanced rectal cancer (LARC), that is cT3-4 N0M0 or anyT N1M0 [2], a multimodality strategy is the best choice with the aims of obtaining better local control and avoiding a destructive surgical treatment, such as anterior resection or abdominoperineal amputation with resection of the sphincter apparatus [4]. A multimodality strategy, developed over recent decades, consists of neoadjuvant concomitant chemoradiotherapy (CRT) followed by more conservative but radical surgery based on total mesorectal excision. In particular, neoadjuvant CRT leads to a reduction in tumour size and stage, increasing resectability and the possibility of sphincter conservation [4–8]. In order to choose the best tailored surgery, assessment of response to neoadjuvant CRT is very important. Evaluation after CRT also has a prognostic impact because of a correlation between histopathological response according to Mandard’s criteria and patient outcome [9–12]. According to the literature, 8–31 % of patients with LARC receiving CRT achieve a pathological complete response (CR) [9, 10, 13–17]. This finding is of great importance especially considering the future possibility of completely avoiding surgery after CRT in these patients [12, 13]. One of the leading candidates to predict histopathological response is 18F-FDG PET/CT, an ever more widely used nuclear medicine technique that detects tumour cells due to their glucose consumption. It is now well established that the biological signal from 18F-FDG PET/CT is important and often more predictive of histological response and outcome than anatomic imaging [18]. Therefore the need to standardize response assessment in PET has become increasingly important and finding semiquantitative and quantitative parameters to evaluate lesion glucose metabolism before and after therapy is becoming a necessity. Wahl et al. were the first in 2009 to propose a draft framework for PET Response Criteria in Solid Tumors (PERCIST) in comparison with anatomic tumour response metrics (WHO criteria and RECIST) [18, 19]. PERCIST can be considered an initial effort to validate quantitative approaches for metabolic treatment response assessment. Many studies in the literature have focused on the evaluation of several FDG PET/CT parameters as predictors of rectal cancer response to CRT, in particular maximum standardized uptake value (SUVmax), visual response assessment (VRA), total lesion glycolysis (TLG) and response index (RI) [12, 14, 15]. The aim of our study was to

Eur J Nucl Med Mol Imaging (2013) 40:853–864

correlate qualitative visual response and various quantification factors from PET scans before and after CRT with the tumour regression grade (TRG) as determined using the system proposed by Mandard et al.

Materials and methods Patients A total of 69 consecutive patients (19 women and 50 men, mean age 59.4 years; see Table 1) referred to our centre for 18 F-FDG PET/CT examinations before neoadjuvant CRT for LARC from January 2008 to December 2011 were retrospectively evaluated. The use of PET for response Table 1 Characteristics of patients Characteristic

Value

Total, n (%) Sex, n (%) Female Male Age (years) Mean±SD Range Quadrant, n (%) 1 2 3 4 4+stenosis Unknown Distance from anal verge (cm) Mean±SD Range Chemoradiotherapy to surgery interval (weeks) Mean±SD

69 (100) 19 (27.5) 50 (72.5) 59.4±11.7 24–80 3 (4.3) 24 (34.8) 17 (24.6) 13 (18.8) 11 (15.9) 1 (1.4) 6.3±2.8 2–12 11.1±2.7

Range 6.6–21.7 PET/CT before CRT to PET/CT after CRT interval (weeks) Mean±SD 15.6±2.8 Range 12.4–29.0 CRT to PET/CT interval (weeks) Mean±SD 6.7±1.8 Range 4.6–17.4 Radiation therapy dose to tumour (Gy) Mean±SD 55.9±0.3 Range 54–56 Radiation therapy dose to nodes (Gy) Mean±SD 49.9±0.7 Range 46–50

Eur J Nucl Med Mol Imaging (2013) 40:853–864

evaluation after neoadjuvant CRT has been routine in our institution since 2005 [11, 14, 16, 17] following several reports published on this topic that demonstrated an improved diagnostic performance (greater sensitivity and specificity) compared with other current techniques, although evidence of an impact on treatment and outcome is still lacking (a potentially appropriate indication) [20, 21]. In our institution some PET parameters before and after neoadjuvant CRT have already been determined (SUVmaxpre, SUVmax-post and change in SUV). This work aimed to investigate the metabolic response also using other parameters including metabolic tumour volume (MTV), TLG, VRA, and PERCIST. We also propose here a new semiquantitative criterion to distinguish metabolic responders from those with residual disease (RD) based on a defined cut-off level (called the “response cut-off”). This cut-off definition is described in detail below. The baseline PET was performed for staging before any treatment. The second PET scan was performed after completion of neoadjuvant CRT (mean time 6.7 weeks, range 4.6 to 17.4 weeks) but before surgery (mean time between CRT and surgery 11.1 weeks). The range of time between CRT and the second PET scan was quite wide; this may be a limitation of this retrospective study. However, if the two longest time intervals (17.4 and 13.1 weeks) are excluded, the time interval range in the other 67 patients was quite small (from 4.6 to 8.1 weeks). PET/CT acquisition protocol All 18F-FDG PET/CT examinations were performed with the same protocol (used in our institution from 2005) and with the same PET/CT hybrid scanner, a Discovery STE (General Electric, Milwaukee, WI), installed in our department. The PET scanner has 24 rings of detectors, subdivided in 8×6 blocks of BGO crystals. The bore has a diameter of 70 cm, with an axial field of view (FOV) extension of 15.7 cm, acquiring 47 slices of thickness 3.3 mm. PET acquisitions were obtained in 3-D modality. The system includes a Lightspeed 16 (General Electric) 16-slice CT scanner. Patients fasted for at least 6 h before the 18F-FDG PET/CT scan and refrained from ingestion of liquid from at least 1 h before until the end of the examination to reduce bladder filling. All patients received an intravenous injection of 2.22 MBq/kg body weight of 18F-FDG (provided by GE Healthcare srl, Milano, Italy). None of the patients showed any adverse events after administration of the radiopharmaceutical. Patients were asked to void their bladder immediately before scanning to minimize the presence of the tracer in the urinary tract. CT images were used for both attenuation correction of emission data (PET) and for morphological usage through image fusion. The acquisition

855

started 60 min after injection and consisted of (a) a scout-view (80 kV, 10 mA, anteroposterior) in order to set the whole-body scan limits (from the orbitomeatal level to the superior portion of the thigh), (b) a lowdose helical CT acquisition (120 kV, 40–100 mA modulated by GE Smart-mA with a noise index of 30, slice thickness 3.75 mm, pitch 1.375, speed 13.75 mm per rotation, rotation time 0.8 s, feet to head) and (c) a PET acquisition (3-D, 3 min per bed position, six or seven bed positions, slice overlap 7, feet-head versus, fully-3D iterative reconstruction including CT-based attenuation correction). 18F-FDG dose, acquisition protocol and reconstruction parameters were chosen on the basis of those used in our previous study [22]. Image interpretation PET data were displayed using commercial software (Xeleris 2.0, GE Healthcare) fused with coregistered noncontrast CT images in coronal, sagittal and transaxial slices. Images were visually interpreted by two experienced nuclear physicians who had no prior knowledge of the results of the pathological analysis of specimens and were then classified into the following categories: CR, partial response (PR), progressive disease (PD), progressive/regressive disease (mixed response), and inconclusive report for images with doubtful findings. Tumour lesions were qualitatively identified as areas of pathologically increased 18F-FDG uptake when other causes of physiological or nonspecific uptake in the intestinal lumen were excluded. In order to quantify the 18F-FDG accumulation, the SUV was calculated was follows: SUV = L(A/bw) where L is the local uptake of 18 F-FDG in the attenuation-corrected PET images (MBq/kg), A is the 18F-FDG activity at the start of PET acquisition (MBq) and bw is the body weight of the patient (kg). Volumes of interest (VOIs) of 1.2 cm diameter were positioned on the area of abnormal 18F-FDG uptake corresponding to the tumour in the baseline scan, then placed matching the position on the scan after CRT. SUVmax was calculated as the maximum SUV within each VOI on the transaxial slices with the highest radioactivity concentration, normalized to the injected dose and the patient’s body weight. The SUVmax values from the baseline scan (SUVmax-pre) and the scan after CRT (SUVmax-post) were also used to assess the tumour response to therapy by calculating the absolute difference δSUVmax: dSUVmax ¼ SUVmax pre  SUVmax post and the percentage decrease in SUV, in the literature commonly called the response index: RI ¼

SUVmax pre  SUVmax post  100% SUVmax pre

856

Eur J Nucl Med Mol Imaging (2013) 40:853–864

Another semiquantitative analysis was performed using PERCIST 1.0, proposed by Wahl et al. in 2009 [18], but classifying the data using PERCIST with a slight modification of, replacing the lean body mass SUV with body weight SUV, as follows: &

& & &

Complete metabolic response (CMR): complete lack of 18 F-FDG uptake within a measurable target lesion such that it was less than mean liver activity and indistinguishable from surrounding background blood-pool levels. Partial metabolic response (PMR): SUVmax decrease of greater than 30 % between the before and after CRT PET scans. Progressive metabolic disease (PMD): SUVmax progression of greater than 30 % between the before and after CRT PET scans. Stable metabolic disease (SMD): no CMR, PMR or PMD.

It is important to point out that the original PERCIST suggested using lean body mass instead of body mass to normalize SUV because it is less dependent on body habitus across the population. However, lean body mass SUV (also called leanSUV or SUL) results may not be reproducible because there is no agreement on a formula for calculating lean body mass [23]. Moreover, in order to assess therapy response, PERCIST 1.0 require comparison of liver and tumour SUL and calculation of the percentage difference in SUL between before and after therapy. In a patient with substantially stable weight, the results are independent of lean body mass. As a consequence, the assessment of response to therapy will not depend on the method of body size measurement, as already reported in the literature [24, 25]. As a consequence, we decided to use body mass instead of lean body mass to normalize all SUVs calculated in this study. Following the definition of measurable target lesion [18], we proposed a response cut-off to distinguish CMR to therapy (CMRt) and RD. CMRt was defined as PET FDG uptake after CMR of ≤1.5×SUVmean_liver +2×SD, and RD was defined as PET FDG uptake after CMR of >1.5×SUVmean_liver +2×SD, where SUVmean_liver is the mean SUV in the liver and SD is its standard deviation, obtained from a region of interest with a diameter of 3 cm, using an Xeleris 2.0 routine workstation. In order to calculate MTV, a fixed threshold value of 40 % of the maximum uptake was used to determine tumour margins automatically using an Advantage workstation (version 4.4; GE Healthcare), according to the previously published method of Larson et al. [26] and Lee [27]. MTV was considered impossible to calculate when FDG uptake was low and diffuse (SUV ≤5.4) and the MTV varied on the basis of the dimensions of the 3-D box area (that is the

volume within which the segmentation was calculated) in the axial plane. In these patients the MTV was taken as zero. In contrast, the MTV was considered quantifiable if two conditions were satisfied: the focal distribution and the MTV segmentation were independent of the change in the 3-D box area dimensions in the axial plane. The image noise, due also to the patient’s weight, quantified by the standard deviation, affected the stability of the volume value within the 3-D box area and therefore the possibility of obtaining a trustworthy MTV. As a consequence, a unique SUVmax threshold to differentiate those in whom MTV could be calculated from those in whom it could not be calculated was not achieved. The SUVmean was then determined as the average of the SUVs in all voxels within the threshold-defined tumour volume. TLG was then calculated as: TLG = SUVmean × MTV. If FDG uptake was normal or near normal, tumour volume was considered impossible to calculate and therefore TLGpost was taken as zero. We also calculated and considered both the arithmetic difference and the percentage change in MTV and TLG from PET scans before and after CRT (δMTV, δTLG, ΔMTV% and ΔTLG%, respectively). The formulae are as follows: dMTV ¼ MTVpre  MTVpost

ΔMTV % ¼

MTVpre  MTVpost  100% MTVpre

dTLG ¼ SUVmeanpre  MTVpre  SUVmeanpost  MTVpost

ΔTLG% ¼

SUVmeanpre  MTVpre  SUVmeanpost  MTVpost  100% SUVmeanpre  MTVpre

Neoadjuvant therapy Neoadjuvant therapy consisted of simultaneous CRT. External radiation therapy was performed as follows: three different pelvic volumes were identified (regional pelvic nodal regions, mesorectal space and neoplastic volume) and treated with 50 Gy, 53 Gy and 56 Gy, respectively, in the same time frame (25 fractions over 33 consecutive days). A concurrent boost was used, and a 2-cm margin was placed around the gross tumour volume for the boost volume. PETdefined pelvic lesions, the biological target volume, were incorporated into the gross tumour volume even if not demonstrated by CT or MRI scans. Patients were positioned prone (12–15°) on an ‘up-down table’ device with a hypogastric compressor, and Vac Fix vacuum immobilization of the legs and pelvis [16, 28, 29].

Eur J Nucl Med Mol Imaging (2013) 40:853–864

During the radiotherapy course patients underwent chemotherapy with 5-fluorouracil administered as a protracted venous infusion for 32–34 days at a daily dose of 300 mg/m2. Surgery All 69 patients underwent surgery after a mean time of 77.5 days (range 46–152 days) after CRT completion. Conservative surgery was possible in 60 patients, while 9 patients needed a permanent colostomy. Tumour distance from the anal verge was 2 to 12 cm (mean 6.3 cm). PET findings alone were never used to change the surgical approach. At staging, all tumour dimensions were >2 cm (mean 5.1 cm, range 2.0–12.0 cm) and the volumes were greater than 4.8 cm3 (mean 19.0 cm3, range 4.7–74.3 cm3), therefore no problems from the partial volume effect occurred. Histopathology All pathological analyses of the surgical specimens were performed by physicians belonging to the same hospital pathology service and read according to the protocol proposed by Quirke and Dixon [30]. Pathological CR after preoperative CRT was defined as absence of cancer cells. The TRG as described by Mandard et al. was used to classify the pathological response to neoadjuvant CRT: TRG 1 complete regression, TRG 2 presence of rare residual cancer cells scattered throughout fibrotic tissue, TRG 3 increased number of residual cancer cells but fibrosis still predominant, TRG 4 residual cancer outgrowing fibrosis, and TRG 5 no regressive changes detectable [31]. We decided to group those with TRG1/2 as responders and those with TRG 3–5 as nonresponders.

857

Also three categorical variables were considered in the analysis, VRA, PERCIST and response cut-off. On the basis of the VRA findings, patients were divided into two groups: those with CR vs. all those without CR (“other than CR”, i.e. those with PR, PD, mixed response or inconclusive report). The patients were also divided into two groups on the basis of the PERCIST data: CMR vs. all those without CMR (“other than CMR”, i.e. those with PMR, PMD and SMD). Pearson’s chi-squared analysis of the VRS, PERCIST and response cut-off cross-tabulated data was performed. Receiver operating characteristic (ROC) analysis was performed for the continuous variables that were found to be significantly correlated with TRG. The area under the curve (AUC) indicates the discriminatory power of a test (a perfect diagnostic test has an AUC of 1.0). We estimated the point with the minimum distance from the upper-left corner of the plot [with coordinate (0,1)], that is the best cut-off value of the predictor maximizing the combination of sensitivity and specificity. All tests were two-sided, with statistical significance taken as p<0.05. All analyses were performed using the SPSS v.17 statistical software package (SPSS, Chicago, IL).

Results Pathology analysis According to the criteria of Mandard et al. [31], the surgical specimens of the 69 patients were classified as TRG 1 in 25 patients (36.2 %), TRG 2 in 24 (34.8 %), TRG 3 in 17 (24.6 %) , and TRG 4 in 3 (4.3 %). There were no patients with TRG 5 (Table 2). As already mentioned, we classified the patients into two groups: responders (TRG 1/2, 49 patients, 71 %) and nonresponders (TRG 3–5, 20 patients, 29 %).

Statistical analysis PET response assessment by SUV The relationships between surgery histological findings (TRG) and possible clinical predictors were investigated using the nonparametric Mann–Whitney test in two independent groups of patients: 49 responders without RD (TRG 1) or only microscopic disease (TRG 2), and 20 nonresponders with TRG 3 or 4. The PET/CT quantitative parameters included in the analysis were: (a) SUVmax-pre, (b) MTV-pre, and (c) TLG-pre values from the baseline PET/CT scan, (d) SUVmax-post, (e) MTV-post, and (f) TLG-post values from the PET/CT scan after CRT, the differences (g) δSUVmax, (h) δMTV and (i) δTLG between the baseline scan and the scan after CRT, and the percentage decreases (l) RI, (m) ΔMTV% and (n) ΔTLG% between baseline scan and the scan after CRT. Univariate binary logistic regression analyses were also performed and the corresponding p values are reported.

The mean baseline SUVmax was 18.0 (range 7.2–57.1, Table 3). The responders (patients with TRG 1 or 2) showed a mean baseline SUVmax of 17.4 (SD 7.5). The nonresponders (TRG 3–5) showed a mean baseline SUVmax of 19.6 (SD 11.5). The difference was not statistically significant (p=0.634). After CRT the mean SUVmax was 5.2 (range 2.2–12.6). In the responders the mean SUVmax was 4.2 (SD 1.4). In the nonresponders the mean SUVmax was 7.7 (SD 2.9). The difference was statistically significant (p<0.001; Fig. 1a). The mean difference between SUVmax-pre and SUVmaxpost was 12.8 (range 0.4–44.5). In the responders the mean δSUVmax was 13.2 (SD 7.6). In the nonresponders the mean δSUVmax was 11.9 (SD 10.4). The difference was not statistically significant (p=0.209). The mean percentage difference

Visual PET response assessment In the VRA 51 patients showed CR, 13 PR, 3 an inconclusive report and 2 a mixed response. Classifying the patients as CR and “other than CR” was significantly predictive of the TRG response groups (TRG 1/2 responders and TRG 3–5 nonresponders) with a sensitivity of 85.7 % and a specificity of 55.0 % (p<0.001; Table 4). PET response assessment by PERCIST In the classification using PERCIST, 4 patients showed CMR, 63 PMR and 2 SMD. None of the patients showed PMD. Classifying the patients as CMR and “other than CMR” was not significantly predictive of the TRG response groups (TRG 1/2 responders and TRG 3–5 nonresponders) with a sensitivity of 100.0 % and a specificity of 8.2 % (p=0.188; Table 5). PET response assessment by response cut-off In the classification using the response cut-off, 31 patients showed CMRt and 38 RD. Classifying the patients as CMRt

0.634 0.379 0.315 <0.001* 0.002* <0.001* 0.209 0.117 0.905 <0.001* 0.006* 0.002* *p < 0.05.

between SUVmax-pre and SUVmax-post was 66.9 % (range 3.9–91.4 %). In the responders the mean RI was 71.9 % (SD 13.1). In the nonresponders the mean RI was 54.7 % (SD 20.3). The difference was statistically significant (p<0.001; Fig. 1d).

SUVmax-pre (g/ml) 18.0±8.8 (7.2 – 57.1) 17.4±7.5 (7.9 – 42.3) 19.6±11.5 (7.2 – 57.1) MTVpre (cm3) 19.0±13.0 (2.5 – 74.3) 18.8±13.8 (4.7 – 74.3) 19.5±11.1 (2.5 – 46.4) TLGpre (g/ml·cm3) 223.9±258.6 (10.7 – 1,585.5) 206.2±217.9 (26.7 – 1,204.2) 267.3±341.6 (10.7 – 1,585.5) After CRT SUVmax-post (g/ml) 5.2±2.5 (2.2 – 12.6) 4.2±1.4 (2.2 – 9.5) 7.7±2.9 (2.9 – 12.6) MTVpost (cm3) 5.6±8.8 (0.0 – 50.6) 3.5±5.8 (0.0 – 24.0) 10.5±12.5 (0.0 – 50.6) TLGpost (g/ml·cm3) 23.2±49.1 (0.0 – 359.0) 10.4±17.5 (0.0 – 63.5) 54.6±80.0 (0.0 – 359.0) Differences between SUVmax-pre – SUVmax-post (g/ml) 12.8±8.5 (0.4 – 44.5) 13.2±7.6 (3.0 – 37.7) 11.9±10.4 (0.4 – 44.5) scans MTVpre – MTVpost (cm3) 13.5±13.6 (−4.2 – 74.3) 15.3±14.9 (−4.2 – 74.3) 9.0±8.3 (−4.2 – 27.6) TLGpre – TLGpost (g/ml·cm3) 200.7±233.3 (10.7 – 1,226.5) 195.8±218.3 (11.0 – 1,204.1) 212.7±272.4 (10.7 – 1,226.5) RI (%) 66.9±17.2 (3.9 – 91.4) 71.9±13.1 (37.5 – 91.4) 54.7±20.3 (3.9 – 84.3) ΔMTV% 68.2±42.2 (−52.4 – 100.0) 74.1±43.0 (−52.4 – 100.0) 53.6±37.1 (−9.1 – 100.0) ΔTLG% 87.7±18.5 (14.2 – 100.0) 91.0±17.0 (14.8 – 100.0) 79.5±20.0 (14.2 – 100.0)

31 (44.9) 38 (55.1)

Non-responders (TRG 3–5)

4 (5.8) 63 (91.3) 2 (2.9) 0 (0)

Responders (TRG 1/2)

13 (18.8) 3 (4.3) 2 (2.9)

All patients

51 (73.9)

Parameter

Partial response Inconclusive report Mixed response PERCIST analysis Complete metabolic response Partial metabolic response Stable metabolic disease Progressive metabolic disease Response cut-off Complete metabolic response to therapy Residual disease

25 (36.2) 24 (34.8) 17 (24.6) 3 (4.3) 0 (0)

PET/CT scan

Tumour regression grade (TRG) 1 2 3 4 5 Visual response assessment (VRA) Complete response

p value

No. (%) of patients (n = 69)

Table 3 Main quantitative parameters from each PET/CT scan in the all patients and in two groups divided according to histological findings. Values are means±SD (range)

Variable

Baseline

Table 2 Categorical variables

0.360 0.832 0.384 <0.001* 0.013* 0.002* 0.551 0.088 0.784 0.001* 0.072 0.033*

Eur J Nucl Med Mol Imaging (2013) 40:853–864 Mann–Whitney Logistic regression

858

Eur J Nucl Med Mol Imaging (2013) 40:853–864

859

Fig. 1 Quantitative factors that were found to be significantly different (Mann–Whitney test) between responders and nonresponders after CRT as shown in Table 3 (a SUVmax, b MTV, c TLG), together with the percentage changes in (d) RI, (e) ΔMTV% and (f) ΔTLG% from the PET/CT scans before and after CRT

Table 4 Cross-tabulation between the two categorical variables in the VRA (CR or “other than CR”) in relation to histological findings (TRG 1/2 responders or TRG 3/4 nonresponders) VRA

TRG

Total p value (chi-squared analysis) 1/2 (responders) 3/4 (nonresponders)

CR 42 “Other 7 than CR” Total 49

9 11

51 18

20

69

<0.001

and RD was significantly predictive of the TRG response groups (TRG 1/2 responders and TRG 3–5 nonresponders) with a sensitivity of 80.0 % and a specificity of 55.1 % (p=0.009; Table 6). PET response assessment by MTV The mean MTV from the baseline PET/CT was 19.0 cm3 (range 2.5–74.3 cm3; Table 3). The responders (TRG 1/2) showed a mean MTV of 18.8 cm3 (SD 13.8 cm3). The nonresponders (TRG 3–5) showed a mean MTV of 19.5 cm3 (SD 11.5 cm3). The difference was not statistically significant (p=0.379, Mann–Whitney analysis). The mean MTV

860

Eur J Nucl Med Mol Imaging (2013) 40:853–864

Table 5 Cross-tabulation between the two categorical variables in the PERCIST classification (CMR or “other than CMR”) in relation to histological findings (TRG 1/2 responders or TRG 3/4 nonresponders) PERCIST TRG Total p value classification (chi-squared analysis) 1/2 (responders) 3/4 (nonresponders) CMR “Other than CMR” Total

4 45

0 20

4 65

49

20

69

0.188

after CRT was 5.6 cm3 (range 0.0–50.6 cm3). The responders showed a mean MTV of 3.5 cm3 (SD 5.8 cm3). The nonresponders showed a mean MTV of 10.5 cm3 (SD 12.5 cm3). The difference was statistically significant (p=0.002, Mann–Whitney analysis; Fig. 1b). The median difference between MTV-pre and MTV-post was 13.5 cm3 (range −4.2 to +74.3 cm3). The responders showed a mean δMTV of 15.3 cm3 (SD 14.9 cm3). The nonresponders showed a mean δMTV of 9.0 cm3 (SD 8.3 cm3). The difference was not statistically significant (p=0.117). The mean percentage change in MTV between before and after CRT was 68.2 % (range −52.4 % to +100 %). The responders showed a mean ΔMTV% of 74.1 % (SD 43.0 %). The nonresponders showed a mean ΔMTV% of 53.6 % (SD 37.1 %). The difference was statistically significant (p=0.006; Fig. 1e). A univariate binary logistic regression was then performed and the corresponding p value was 0.072 (not significant). MTV could not be calculated in 32 patients. In some patients, even those with a low SUVmax (≥ 3.7), the MTV was quantifiable. The criteria for definition are discussed in the section Materials and methods. This challenging matter is examined further the Discussion. PET response assessment by TLG The mean TLG from the baseline PET/CT was 223.9 cm3 (range 10.7–1585.5 cm3; Table 3). The responders showed a mean TLG value of 206.2 cm 3 (SD 217.9 cm 3). The

Table 6 Cross-tabulation between the two categorical variables in response cut-off classification (CMRt and RD) in relation to histological findings (TRG 1/2 responders or TRG 3/4 nonresponders) Response TRG Total p value cut-off (chi-squared classification 1/2 (responders) 3/4 (nonresponders) analysis) CMRt RD

27 22

4 16

31 38

Total

49

20

69

0.009

nonresponders showed a mean TLG value of 267 cm3 (SD 341.6 cm3). The difference was not statistically significant (p=0.315, Mann–Whitney analysis). The mean TLG after CRT was 23.2 cm3 (range 0.0–359.0 cm3). The of responders showed a mean TLG value of 10.4 cm3 (SD of 17.5 cm3). The nonresponders showed a mean TLG value of 54.6 cm3 (SD of 80.0 cm3). The difference was statistically significant (p<0.001, Mann–Whitney analysis; Fig. 1c). The median difference between TLG-pre and TLG-post was 200.7 cm3 (range 10.7 to 1226.5 cm3). The responders showed a mean δTLG value of 195.8 cm3 (SD 218.3 cm3). The nonresponders showed a mean δTLG value of 212.7 cm3 (SD 272.4 cm3). The difference was not statistically significant (p=0.905). The mean percentage change in TLG between before and after CRT was 87.7 % (range 14.2 to 100 %). The responders showed a mean ΔTLG% of 91.0 % (SD 17.0 %). The nonresponders showed a mean ΔTLG% of 79.5 % (SD 20.0 %). The difference was statistically significant (p=0.002; Fig. 1f). ROC analysis ROC analysis was performed for the six continuous variables that were found to be significantly related to TRG (SUVmax-post, MTV-post, TLG-post, RI, ΔMTV% and ΔTLG%; Table 7). The SUVmax-post AUC was 0.846, with a sensitivity of 85.7 % and a specificity of 80 % for a cut-off of 5.1. The MTV-post AUC was 0.722, with a sensitivity of 65.3 % and a specificity of 80 % with a cut-off of 2.1 cm3. The TLGpost AUC was 0.790, with a sensitivity of 85.7 % and a specificity of 75 % for a cut-off of 23.4 cm3. The RI AUC was 0.772, with a sensitivity of 83.7 % and a specificity of 70 % for a cut-off of 61.8 %. The ΔMTV% AUC was 0.695, with a sensitivity of 69.4 % and a specificity of 80 % for a cut-off of 81.4 %. The ΔTLG% AUC was 0.726, with a sensitivity of 69.4 % and a specificity of 80 % for a cut-off of 94.2 % (Figs. 2 and 3).

Discussion Preoperative CRT for LARC is widely considered the best approach to preventing local recurrence. Moreover, neoadjuvant CRT gives the chance to avoid colostomy in distal rectal carcinoma, with the obvious better quality of life [10]. For these reason, precise restaging to assess the outcome of the preoperative treatment is essential. MRI and rectal US seem to have better diagnostic properties than CT in rectal cancer [32], but CT is still routinely requested by surgeons for local staging. Depending on the difficulty in differentiating persistent disease from scar tissue, morphological imaging tends to upstage the disease in the work-up after CRT, and the accuracy

Eur J Nucl Med Mol Imaging (2013) 40:853–864 Table 7 ROC analysis for the six prognostic factors SUVmax, MTV and TLG in after CRT and the percentage change in SUVmax (RI), ΔMTV% and ΔTLG% between before and after CRT

861

Factor

AUC

AUC p value

SUVmax-post MTV-post TLG-post RI ΔMTV% ΔTLG%

0.846 0.722 0.790 0.772 0.695 0.726

<0.001 0.004 <0.001 <0.001 0.012 0.003

Cut-off 5.1 2.1 23.4 61.8 81.4 94.2

g/ml cm3 g/ml·cm3 % % %

Sensitivity (%)

Specificity (%)

85.7 65.3 85.7 83.7 69.4 69.4

80.0 80.0 75.0 70.0 80.0 80.0

of MRI and CT seems to be lower than that of FDG PET/CT [33]. A large number of studies have demonstrated a relatively strong relationship between 18F-FDG uptake and cancer cell numbers [34]. As a result, it is reasonable to expect that the loss of viable cancer cells would be accompanied by a decrease in tumour 18F-FDG uptake. Moreover, the inability of 18 F-FDG to discriminate minimal tumour burden and no tumour burden has also been clearly shown [18]. The superiority of 18F-FDG PET for monitoring therapeutic efficacy after or during chemotherapy and/or CRT in a wide range of clinical settings has been demonstrated or suggested in several studies [20] and 18F-FDG PET is considered potentially appropriate for evaluating the response of LARC to preoperative CRT [21]. In spite of these efforts to investigate the role of PET as a predictive tool before surgery in patients with LARC, in the last version of the National Comprehensive Cancer Network (NCCN) Guidelines on Rectal Cancer [35], in the paragraph on response to neoadjuvant treatment, PET is not even mentioned. Recent studies on this topic are summarized in Table 8.

In this study eight PET parameters, both qualitative and quantitative, were found to be significantly related to pathological response to neoadjuvant CRT: VRA, reporting by response cut-off, SUVmax-post, MTVpost, TLGpost, RI, ΔMTV% and ΔTLG%. SUVmax-post demonstrated the highest AUC, sensitivity and specificity. Between the two systems for reporting, VRA showed better accuracy and statistical significance than the system based on response cut-off. In our PET centre, we point out RI in the report conclusion as a helpful tool for radiotherapists and surgeons in the therapeutic management of the patient. We therefore suggest that the qualitative approach to the PET examination should be trusted, and that RI and SUVmax-post cut-off values (61.8 % and 5.1, respectively) should be routinely used in clinical practice. The high prognostic value of PET could be used to avoid wide tumour resection and therefore a possible permanent colostomy in patients with low rectal cancer. However, this requires investigation in more depth in prospective studies.

Fig. 2 ROC curves for the continuous variables found to be significantly related TRG after CRT: SUVmax-post, MTVpost and TLGpost

Fig. 3 ROC curves for the continuous variables that were found to show significant changes after CRT: RI, ΔMTV% and ΔTLG%

862

Eur J Nucl Med Mol Imaging (2013) 40:853–864

Table 8 Recent papers studies of the use of PET in LARC Reference

No. of patients

PET parameter

Sensitivity (%)

Specificity (%)

Cut-off

Take home message

[15]

21

RI

86

85

75 %

[37]

30

SUVmax-post

77

89

4.4

[38]

15

RI ΔTLG

70 90

100 80

62.5 % 69.5 %

[39]

30

Visual analysis

45

75

[14]

87

RI

85

80

65 %

[40]

80

SUVmax-post RI

88 94

34 31

5.0 g/ml 66 %

[41]

70

[42]

35

58 60 93

78 84 19

4.0 g/ml 63 % 64 %

Present study

69

SUVmax-post RI ΔSUVmax (SUVmax-pre/ SUVmax-post) SUVmax-post MTVpost TLGpost RI ΔMTV% ΔTLG% VRA

86 65 86 84 69 69 55

80 80 75 70 80 80 86

5.1 2.1 cm3 23.4 cm3 61.8 % 81.4 % 94.2 %

Tumour downstaging and CR are associated with greater RI SUVmax and MRI apparent diffusion coefficient are the best parameters to define response to treatment, by differentiating fibrosis from viable tumour tissue RI and ΔTLG cut-off are the best predictors of no-evidence-of-disease status and freedom from recurrence PET 7 weeks after CRT is not able to predict the pathological response. Criticality: liver uptake was considered as cut-off for visual analysis RI seems the best predictor to identify response to CRT PET/CT supplies limited predictive information. However, SUVmax-post cut-off appears to be associated with a favourable patient outcome Posttreatment SUV and RI, and greater time from CRT to surgery correlate with pathological CR SUVmax, ΔSUVmax, MTV and ΔMTV (calculated with different cut-off values) are not correlated with TRG Among eight statistically significant PET parameters, SUVmax-post and VRA showed the best accuracy in predicting TRG

Response cut-off

80

55

Since we found that PERCIST evaluations were not significantly correlated with TRG (p =0.188, sensitivity 100 % and specificity 8.2 %) a deeper investigation of this qualitative approach was performed. CMR was not significantly predictive of TRG (high sensitivity but very low specificity), and this is particularly evident in our routine clinical experience. This is in contrast to other malignancies such as lymphoma, in which a CMR is associated with FDG uptake that is clearly not pathological, and rectal carcinoma, in which in a patient with CMR, FDG uptake within the tumour site after neoadjuvant CRT may be higher than the surrounding background blood-pool levels. There may be several reasons for this metabolic behaviour, such as postactinic inflammation, or physiological tracer washout via the intestine, etc. These considerations lead to the possibility that a single definition of RD after therapy may not be valid for every type of tumour. This issue led us to define a new cut-off to assess metabolic response in rectal cancer (response cut-off). We found a statistically significant correlation between this classification and TRG (p=0.009) and we

advise that the usefulness of this new predictive criterion be evaluated in further studies in other solid tumours. A subject of prolonged discussion among our team of physicians and physicists concerns the calculation of the RD volume in the presence of low SUVmax-post values. In fact, when SUVmax is less than 5.5, the segmentation with a threshold of 40 % of SUVmax will give a large VOI that also includes areas with SUVmax of about 2.2, commonly considered as physiological tissue uptake. In our study, a SUVmaxpost of ≤5.5 was seen in 52 patients (75 % of our patients). Certainly, both maximum liver and normal intestinal uptake (which is constantly below SUVmax 5.5), as proposed by some authors, are not acceptable as thresholds for RD segmentation. This problem was known as early as 1997: Erdi and colleagues [36] suggested that only if the tumour to background ratio is greater than 3:1 can the area be accurately determined on a tomographic section of the tumour. In this case, if mediastinum FDG uptake could be considered as background activity (commonly around SUVmean 2.0), subsequently all MTVs with SUVmax≤6.0 (resulted by 3•2.0)

Eur J Nucl Med Mol Imaging (2013) 40:853–864

should not be drown. Our proposed SUVmax-post prognostic cut-off value of ≤5.1 is not far from 6.0. However, for segmentation we decided to consider not only the pure SUV value but also the distribution of FDG within the rectal wall (focal vs. diffuse) and that MTV segmentation was independent of or dependent on the changes in the 3-D box area dimensions. In our experience, this approach allows the MTV also to be drawn with low SUVmax (≥3.7). It has to be noted that the SUVmax-post cut-off value of 5.1, proposed here, is close to the SUVmax values of 5.5 and 6.0, as debated above. In our opinion, this similarity lends weight to the use of this cut-off value. Nevertheless, some tricky questions remain open: What value of SUVmax should be considered as the cut-off to decide if the MTV should be drawn or not? Could the FDG distribution (focal vs. diffuse) be useful for determining if the MTV has to be calculated or not? Is the commonly used threshold of 40 % for segmentation incorrect when SUVmax is low? Could the 3:1 tumour to background ratio be an accurate method for segmentation of the MTV? In short, is it more trouble than it’s worth. This discussion leads to some considerations: 1. Eight PET parameters are predictive of pathological response. This means that the 18F-FDG PET/CT scan is able to accurately stratify patients with LARC whatever the method chosen to interpreted the images. 2. Among many PET parameters, some of which cannot immediately be obtained (such as MTV and TLG) from a routine workstation used for reporting and are therefore of little impact in routine practice, the most commonly used (SUVmax-post and VRA) showed the best accuracy in predicting TRG. 3. The derived SUVmax-post and RI cut-off values of 5.1 and 61.8 %, respectively, should be seriously considered for use in clinical practice as helpful tools for reporting. 4. Since a statistically significant correlation between the newly proposed response cut-off and TRG was found, we advise further investigations applying this criterion in other solid tumours.

863

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

Conflicts of interest None. 15.

References 16. 1. Delbeke D, Walker R. Colorectal Cancer”. In: Delbeke D, Israel O, editors. Hybrid PET/CT and SPECT/CT imaging. Springer; 2010. p. 261–92. 2. Gil-delgado MA, Khayat D. Cancro del colon e del retto. In: Pollock RE, Doroshok JH, Khayat D, Nakao A, O’Sullivah B,

17.

editors. UICC manuale di oncologia clinica. Minerva medica; 2008, VIII ed. p. 511–29. Bosset JF, Collette L, Calais G, Mineur L, Maingon P, RadosevicJelic L, et al. EORTC Radiotherapy Group Trial 22921. Chemotherapy with preoperative radiotherapy in rectal cancer. N Engl J Med. 2006;355(11):1114–23. Erratum in: N Engl J Med. 2007;357(7):728. De Paoli A, Chiara S, Luppi G, Friso ML, Beretta GD, Del Prete S, et al. Capecitabine in combination with preoperative radiation therapy in locally advanced, resectable, rectal cancer: a multicentric phase II study. Ann Oncol. 2006;17(2):246–51. Graf W, Dahlberg M, Osman MM, Holmberg L, Pählman L, Glimelius B. Short-term preoperative radiotherapy results in downstaging of rectal cancer: a study of 1316 patients. Radiother Oncol. 1997;43(2):133–7. Rödel C, Martus P, Papadoupolos T, Füzesi L, Klimpfinger M, Fietkau R, et al. Prognostic significance of tumor regression after preoperative chemoradiotherapy for rectal cancer. J Clin Oncol. 2005;23(34):8688–96. Rosenberg R, Nekarda H, Zimmermann F, Becker K, Lordick F, Hofler H, et al. Histopathological response after preoperative radiochemotherapy in rectal carcinoma is associated with improved overall survival. J Surg Oncol. 2008;97(1):8–13. Medich D, McGinty J, Parda D, Karlovits S, Davis C, Caushaj P, et al. Preoperative chemoradiotherapy and radical surgery for locally advanced distal rectal adenocarcinoma: pathologic findings and clinical implications. Dis Colon Rectum. 2001;44(8):1123–8. Habr-Gama A, Perez RO, Nadalin W, Nahas SC, Ribeiro Jr U, Silva E, et al. Long-term results of preoperative chemoradiation for distal rectal cancer correlation between final stage and survival. J Gastrointest Surg. 2005;9(1):90–9. Valentini V, Coco C, Cellini N, Picciocchi A, Fares MC, Rosetto ME, et al. Ten years of preoperative chemoradiation for extraperitoneal T3 rectal cancer: acute toxicity, tumor response, and sphincter preservation in three consecutive studies. Int J Radiat Oncol Biol Phys. 2001;51(2):371–83. Capirci C, Rubello D, Chierichetti F, Crepaldi G, Fanti S, Mandoliti G, et al. Long-term prognostic value of 18F-FDG PET in patients with locally advanced rectal cancer previously treated with neoadjuvant radiochemotherapy. AJR Am J Roentgenol. 2006;187(2): W202–8. Huh JW, Min JJ, Lee JH, Kim HR, Kim YJ. The predictive role of sequential FDG-PET/CT in response of locally advanced rectal cancer to neoadjuvant chemoradiation. Am J Clin Oncol. 2012;35(4):340–4. Habr-Gama A, Perez RO, Nadalin W, Sabbaga J, Ribeiro Jr U, Silva e Sousa Jr AH, et al. Operative versus nonoperative treatment for stage 0 distal rectal cancer following chemoradiation therapy: long-term results. Ann Surg. 2004;240(4):711–7. Capirci C, Rubello D, Pasini F, Galeotti F, Bianchini E, Del Favero G, et al. The role of dual-time combined 18-fluorodeoxyglucose positron emission tomography and computed tomography in the staging and restaging workup of locally advanced rectal cancer, treated with preoperative chemoradiation therapy and radical surgery. Int J Radiat Oncol Biol Phys. 2009;74(5):1461–9. Melton GB, Lavely WC, Jacene HA, Schulick RD, Choti MA, Wahl RL, et al. Efficacy of preoperative combined 18-fluorodeoxyglucose positron emission tomography and computed tomography for assessing primary rectal cancer response to neoadjuvant therapy. J Gastrointest Surg. 2007;11(8):961–9. Capirci C, Rampin L, Erba PA, Galeotti F, Crepaldi G, Banti E, et al. Sequential FDG-PET/CT reliably predicts response of locally advanced rectal cancer to neo-adjuvant chemo-radiation therapy. Eur J Nucl Med Mol Imaging. 2007;34(10):1583–93. Capirci C, Rubello D, Chierichetti F, Crepaldi G, Carpi A, Nicolini A, et al. Restaging after neoadjuvant chemoradiotherapy for rectal

864

18.

19.

20.

21. 22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

Eur J Nucl Med Mol Imaging (2013) 40:853–864 adenocarcinoma: role of F18-FDG PET. Biomed Pharmacother. 2004;58(8):451–7. Wahl RL, Jacene H, Kasamon Y, Lodge MA. From RECIST to PERCIST: evolving considerations for PET response criteria in solid tumors. J Nucl Med. 2009;50 Suppl 1:122S–50. Costelloe CM, Chuang HH, Madewell JE, Ueno NT. Cancer response criteria and bone metastases: RECIST 1.1, MDA and PERCIST. J Cancer. 2010;1:80–92. International Atomic Energy Agency. The role of PET/CT in radiation treatment planning for cancer patient treatment. Vienna: IAEA; 2008. International Atomic Energy Agency. Appropriate use of FDGPET for the management of cancer patients. Vienna: IAEA; 2010. Ferretti A, Bellan E, Gava M, Chondrogiannis S, Massaro A, Nibale O, et al. Phantom study of the impact of reconstruction parameters on the detection of mini- and micro-volume lesions with a low-dose PET/CT acquisition protocol. Eur J Radiol. 2012;81(11):3363–70. Erselcan T, Turgut B, Dogan D, Ozdemir S. Lean body mass-based standardized uptake value, derived from a predictive equation, might be misleading in PET studies. Eur J Nucl Med Mol Imaging. 2002;29:1630–8. Paquet N, Albert A, Foidart J, Hustinx R. Within patient variability of 18F-FDG: standardized uptake values in normal tissues. J Nucl Med. 2004;45:784–8. Stahl A, Ott K, Schwaiger M, Weber WA. Comparison of different SUV-based methods for monitoring cytotoxic therapy with FDG PET. Eur J Nucl Med Mol Imaging. 2004;31:1471–8. Larson SM, Erdi Y, Akhurst T, Mazumdar M, Macapinlac HA, Finn RD, et al. Tumor treatment response based on visual and quantitative changes in global tumor glycolysis using PET-FDG imaging. The visual response score and the change in total lesion glycolysis. Clin Positron Imaging. 1999;2(3):159–71. Lee JA. Segmentation of positron emission tomography images: some recommendations for target delineation in radiation oncology. Radiother Oncol. 2010;96(3):302–7. Capirci C, Valvo F, Salviato S. Concurrent boost radiotherapy as preoperative treatment for locally advanced rectal carcinoma: a new beam arrangement. Tumori. 2002;88:325–30. Capirci C, Polico C, Mandoliti G. Dislocation of small bowel volume within box pelvic treatment fields, using the new “Up Down Table” device. Int J Radiat Oncol Biol Phys. 2002;88:325–30. Quirke P, Dixon MF. The prediction of local recurrence in rectal adenocarcinoma by histopathological examination. Int J Colon Dis. 1988;3:127–31. Mandard AM, Dalibard F, Mandard JC, Marnay J, Henry-Amar M, Petiot JF, et al. Pathologic assessment of tumor regression after preoperative CRT of esophageal carcinoma. Clinicopathologic correlations. Cancer. 1994;73:2680–6.

32. Augestad KM, Lindsetmo RO, Stulberg J, Reynolds H, Senagore A, Champagne B, et al. International preoperative rectal cancer management: staging, neoadjuvant treatment, and impact of multidisciplinary teams. World J Surg. 2010;34:2689– 700. 33. Deneke T, Rau B, Hoffmann KT. Comparison of CT, MRI and FDG-PET in response prediction of patients with locally advanced rectal cancer after multimodal preoperative therapy: is there a benefit in using functional imaging? Eur Radiol. 2005;15:1658–66. 34. Bos R, van Der Hoeven JJ, van Der Wall E, van Der Groep P, van Diest PJ, Comans EF, et al. Biologic correlates of 18fluorodeoxyglucose uptake in human breast cancer measured by positron emission tomography. J Clin Oncol. 2002;20:379–87. 35. National Comprehensive Cancer Network. Clinical practice guidelines in oncology. Rectal cancer. Table of Contents Discussion, Version 3.2013, table MS-16, www.nccn.org. Accessed 1 Sep 2012. 36. Erdi YE, Mawlawi O, Larson SM, Imbriaco M, Yeung H, Finn R, et al. Segmentation of lung lesion volume by adaptive PET image thresholding. Cancer. 1997;80 Suppl 12:2505–9. 37. Ippolito D, Monguzzi L, Guerra L, Deponti E, Gardani G, Messa C, et al. Response to neoadjuvant therapy in locally advanced rectal cancer: assessment with diffusion-weighted MR imaging and 18FDG PET/CT. Abdom Imaging. 2012;37:1032–40. 38. Guillem JG, Moore HG, Akhurst T, Klimstra DS, Ruo L, Mazumdar M, et al. Sequential preoperative fluorodeoxyglucosepositron emission tomography assessment of response to preoperative chemoradiation: a means for determining longterm outcomes of rectal cancer. J Am Coll Surg. 2004;199(1):1–7. 39. Kristiansen C, Loft A, Berthelsen AK, Graff J, Lindebjerg J, Bisgaard C, et al. PET/CT and histopathologic response to preoperative chemoradiation therapy in locally advanced rectal cancer. Dis Colon Rectum. 2008;51(1):21–5. 40. Martoni AA, Di Fabio F, Pinto C, Castellucci P, Pini S, Ceccarelli C, et al. Prospective study on the FDG-PET/CT predictive and prognostic values in patients treated with neoadjuvant chemoradiation therapy and radical surgery for locally advanced rectal cancer. Ann Oncol. 2011;22(3):650–6. 41. Shanmugan S, Arrangoiz R, Nitzkorski JR, Yu JQ, Li T, Cooper H, et al. Predicting pathological response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer using 18FDG-PET/CT. Ann Surg Oncol. 2012;19(7):2178–85. 42. Chennupati SK, Quon A, Kamaya A, Pai RK, La T, Krakow TE, et al. Positron emission tomography for predicting pathologic response after neoadjuvant chemoradiotherapy for locally advanced rectal cancer. Am J Clin Oncol. 2012;35(4):334–9.