Evaluation of ¹⁸F-FDG-PET for Detection of Suboptimal Response of Rectal Cancer to Preoperative Chemoradiotherapy: A Prospective Analysis

Abstract

Background:

Early Identification of inadequate response to preoperative chemoradiotherapy (CRT) may spare rectal cancer patients the toxicity of ineffective treatment. We prospectively evaluated tumor response with ¹⁸F- fluorodeoxyglucose (FDG) positron emission tomography (PET) early in the course of preoperative CRT.

Methods:

Twenty-seven prospectively accrued patients with locally advanced rectal cancer (T_3–4/N₁) received preoperative CRT (5040 cGy + 5FU-based chemotherapy). Patients underwent PET scanning before and 8–14 days after commencement of CRT. Scans were interpreted using three standard parameters: SUV_max, SUV_average, total lesion glycolysis (TLG); and an investigational parameter: visual response score (VRS). Percent pathologic response was quantified as a continuous variable. All PET parameters were correlated with pathology. Pathologic complete/near-complete response was defined as ≥95% tumor destruction, suboptimal response as <95%. Statistical analysis was performed using the Wilcoxon rank sum test and receiver operating characteristic (ROC) curve analysis.

Results:

Eleven patients (41%) had pathologic complete/near-complete response; 16 (59%) had suboptimal response. SUV_max, SUV_average, and TLG did not discriminate between responders and non-responders. VRS was statistically significantly higher for complete/near-complete responders than for suboptimal responders (65% vs. 33%, p=0.02). Suboptimal responders were identified with 94% sensitivity and 78% accuracy using a VRS cut-off of 50%.

Conclusions:

In this pilot study, FDG-PET at 8–14 days after the beginning of preoperative CRT was unsuccessful at predicting oncologic outcome with enough accuracy to justify an early change in therapy.

INTRODUCTION

Preoperative chemoradiotherapy (CRT), followed by radical surgery performed according to the principles of total mesorectal excision (TME), has emerged as the preferred treatment for locally advanced (endorectal ultrasound stage T3–4 and/or N1, or clinically bulky) rectal cancer.^[1] Eight to 31 percent of patients treated with this approach achieve pathologic complete response (pCR).^[1–7] Data from our institution and others suggest that patients who achieve near-complete or complete (≥95%) response experience improved long-term local control, as well as better overall and recurrence-free survival.^{[3, 8–12]}

Early, accurate determination of response to preoperative CRT is important not only for prognostic purposes, but because it also enables the clinician to alter standard 5FU-based preoperative therapy in patients experiencing poor initial response. (Some alternative regimens utilizing other chemotherapeutic agents such as oxaliplatin, irinotecan, cisplatin, mitomycin C and raltitrexed have shown promising results in the preoperative setting^[1–8]; in fact, Gambacorta et al (2004) reported that as many as 39% of patients receiving oxaliplatin obtained a pathologic complete response.^[4] The results of phase I/II trials using molecular antibodies such as cetuximab and bevacizumab preoperatively have also been promising.^[9–11])

A number of studies have demonstrated that early prediction of treatment response is feasible in ovarian, non-small-cell lung, esophageal, and gastric cancers, as well as soft tissue sarcomas.^[13–19] In rectal cancer, changing the preoperative CRT regimen of suboptimal pathologic responders may have the additional benefit of improving these patients’ pathologic responses, positively impacting rates of sphincter preservation and survival. Some studies additionally suggest that local surgical procedures,^[20–22] or even observation alone,^[23] may be viable options for patients demonstrating an excellent response to preoperative CRT. However, this approach requires precise determination of the true extent of pathologic response.

Standard imaging modalities, including endorectal ultrasound (ERUS), computed tomography (CT), and magnetic resonance imaging (MRI) are inadequate in determining rectal cancer response to preoperative CRT.^[24–27] We have previously demonstrated that changes in rectal cancer metabolism, based on pre- and post-CRT imaging with ¹⁸F-fluorodexyglucose (FDG) positron emission tomography (PET), correlate accurately with extent of pathologic response^[28] and may predict long-term oncologic outcome.^[29] However, routine FDG-PET scanning before CRT is not currently considered part of the standard of care for rectal cancer patients. The aim of this study was to prospectively evaluate changes, in different ¹⁸F-FDG-PET parameters, between scans performed prior to and 8–14 days after the commencement of preoperative CRT for patients with locally advanced rectal cancer, in order to identify those with suboptimal response who may benefit from a change in preoperative regimen.

METHODS

Patient Population

We prospectively accrued 27 patients with advanced (ERUS stage T_3–4 and/or N₁), biopsy-proven rectal adenocarcinoma, treated with preoperative CRT and radical rectal resection at Memorial Sloan-Kettering Cancer Center between 2001 and 2005. (Three patients with metastatic disease were included because they received preoperative CRT and underwent resection.)

We stratified patients into two clinically significant groups in order to evaluate the correlation of ¹⁸F-FDG-PET parameters to pathologic response: pathologic complete or near-complete responders [defined as pathologic response ≥ 95%; n=11 (41%)]; and suboptimal responders [defined as pathologic response < 95%; n=16 (59%)]. We used ≥ 95% pathologic response as a cut-off to distinguish between complete/near-complete pathologic responders and suboptimal responders, as we have previously demonstrated that 95% and 100% responders have a similarly favorable outcome.^[8]

This study was approved by the Memorial Sloan-Kettering Cancer Center Institutional Review Board. Each patient signed a written informed consent prior to enrollment.

¹⁸F-FDG PET scans

¹⁸F-FDG-PET was performed with the GE Advance (General Electric Medical Systems, GEMS, Milwaukee, WI) whole body PET-scanner, not earlier than 45 minutes after injection of 10–15 mCi of ¹⁸F-FDG. Prior to beginning preoperative CRT, each patient had an ¹⁸F-FDG-PET scan (baseline scan) (Figure 1). Eight to 14 days (median: 10 days) following commencement of CRT, each patient had a second ¹⁸F-FDG-PET scan (intermediate scan) (Figure 1). All scans were independently interpreted by two dedicated Nuclear Medicine physicians (TA and HY), who were blinded to clinical data. The ¹⁸F-FDG-PET scans were interpreted with particular emphasis on standardized uptake value (SUV_max, SUV_avg), total lesion glycolysis (TLG), and visual response score (VRS). Following individual interpretation, the Nuclear Medicine physicians met to determine a consensus reading, which provided the data for this study. Consensus readings of the ¹⁸F-FDG-PET and the PET parameters of SUV_max, SUV_avg, TLG, and VRS were prospectively recorded on study-specific forms.

Figure 1: Example of ¹⁸F-FDG PET scan prior to (baseline) and 10 days after (intermediate) commencement of preoperative CRT.

The image to the left shows a baseline ¹⁸F-FDG-PET in a patient with rectal cancer (arrow indicates the rectal cancer).

The image to the right shows an intermediate ¹⁸F-FDG-PET in the same patient performed 10 days after commencement of CRT.

(¹⁸F-FDG-PET, ¹⁸F-fluorodeoxyglucose positron emission tomography; CRT, chemoradiotherapy)

Specific PET Parameters

We evaluated the ability of changes in four specific PET parameters to predict ultimate pathologic response. SUV is a calculation representing the uptake of ¹⁸F-FDG by tumor, normalized for dose administered and patient weight. We evaluated both SUV_max, which represents maximum SUV in the region of interest on ¹⁸F-FDG scan, and SUV_avg, which represents average SUV throughout the entire region of interest. TLG represents the summed metabolic rate of the tumor. For each patient, change in the above-mentioned PET parameters (ΔSUV_max, ΔSUV_avg and ΔTLG) between scans was calculated and recorded in percent terms. VRS is a visual graded global assessment of response, recorded as a percentage decrease in tumor uptake between PET scans.

Preoperative Chemoradiotherapy

Following clinical assessment, all patients underwent a course of preoperative CRT consisting of external beam radiation therapy (median dose: 5040 cGy; range: 4860 to 5400 cGy) according to previously published techniques.^[2] A multiple field (3-field) technique was used. The perineum was blocked as much as possible in the lateral fields.

Patients received preoperative 5FU-based chemotherapy concurrently with external beam radiation therapy (EBRT) [bolus, n=2 (7.4 percent); continuous infusion, n=25 (92.6 percent)]. The regimen for bolus chemotherapy consisted of 5FU (325 mg/m²/day) with leucovorin (20 mg/m²/day), given for two cycles of five consecutive days on week one (days 1–5) and five (days 29–33) of radiation therapy. The most common regimen for continuous infusion chemotherapy was 5FU (225 mg/m²/day) for a six-week continuous cycle. Two patients in the continuous infusion chemotherapy group received cetuximab [400mg/m² IV (over 120 minutes) on day 1, followed by weekly doses of 250 mg/m² IV (over 60 minutes); following completion of radiochemotherapy, cetuximab was given alone weekly for an additional 4 weeks prior to surgery] in addition to 5FU.

Radical Rectal Resection

After a median interval of 44 days from completion of preoperative CRT (range: 37–55 days), patients underwent radical rectal resection according to the principles of total mesorectal excision (TME). Briefly, TME includes sharp rectal mobilization in the avascular areolar tissue between the visceral fascia covering the rectum and mesorectum and the parietal fascia covering the pelvic sidewall. TME was performed for middle and low rectal cancers (<10 cm from the anal verge). For high rectal cancers (>10 cm from the anal verge), the mesorectum was transected at a right angle 5 cm distal to the distal-most mucosal edge of the tumor.

Pathologic Assessment

Whole mount pathologic analysis was performed as previously described by Guillem et al.^[30] Each specimen was comprehensively evaluated by a single pathologist (JS), who was blinded to the ¹⁸F-FDG-PET results. As reported in earlier studies from our institution, areas of tumor response were characterized by the replacement of neoplastic glands with loosely collagenized fibrous tissue and scattered, chronically inflammatory cells.^[9,28] Rectal cancer response was recorded as a percentage ranging from no evidence of response (0% response) to pCR (100% response, defined as no viable tumor in the resected specimen).

Statistical Analysis

For each of the PET parameters described above, the percentage change from baseline to intermediate scan was calculated. The association between the changes and final pathologic tumor response was evaluated using the Wilcoxon rank-sum test. A p-value of <0.05 was considered statistically significant. A receiver operating characteristic (ROC) curve was plotted for VRS to identify the cut-off value with the highest accuracy for predicting pathologic suboptimal response. Sensitivity was defined as the proportion of patients with a suboptimal response identified by pathology who had a suboptimal response determined by ¹⁸F-FDG-PET. Accuracy was defined as the proportion of all patients that had a correct determination of response (either suboptimal or complete/near-complete) on ¹⁸F-FDG-PET.

RESULTS

Patient Population

Median age at surgery was 50.5 years (range: 36.1 to 72.5 years). There were 9 females (33%) and 18 males (67%). Sphincter preservation with low anterior resection (LAR) was achieved in 23 patients (85%), while 4 (15%) underwent abdominoperineal resection (APR). Percent pathologic response ranged from 30–100%.

¹⁸F-FDG PET Parameters for Entire Study Population

On baseline scan, median SUV_max was 12.4 (range: 5.9 to 29.2), median SUV_avg was 6.0 (range: 2.9 to 15.0), and median TLG was 133.2 (range: 41.9 to 344.0). By definition, there is no reading for VRS on baseline scan because a prerequisite for VRS is treatment response. On intermediate scan, median SUV_max was 8.6 (range: 2.9 to 14.3), median SUV_avg was 4.0 (range: 0.9 to 10.1), median TLG was 71.9 (range: 7.4 to 237.9), and median VRS was 40% (range: −8% to 80%). The level of agreement between two PET scan readers for VRS reached a concordance correlation coefficient of 0.72, indicating fair agreement beyond chance. Median ΔSUV_max from baseline to intermediate scan was 27% (range: −30.5% to 73.8%), median ΔSUV_avg was 33% (range: −37.9% to 78.5%), and median ΔTLG was 39% (range: −20.7% to 91.4%). The change in ΔSUV_max and ΔSUV_avg were not statistically significant.

Correlation of ¹⁸F-FDG PET Parameters with Pathologic Response and Outcome

The correlation of PET parameters with both groups is shown in Table 1. Only the VRS showed a statistically significant difference between pathologic complete/near-complete responders and suboptimal responders. ROC analysis (Figure 2) revealed that a VRS cut-off of 50% provided the highest sensitivity (94%) and accuracy (78%) for identifying suboptimal pathologic responders. With this cut-off, 15 of 16 suboptimal responders and 6 of 11 complete or near-complete responders were correctly identified. After a median follow-up of 40 months, all patients with a VRS > 50% were alive (n=7, 26%), and none had developed recurrent disease; 6 were pathologic responders and 1 was a non-responder. In patients with a VRS ≤ 50%, 3 had developed recurrence and 1 had died of disease. Description of patients according to T-stage and N-stage is shown in Table 2. The level of agreement between both PET readers assessing VRS is shown in Figure 3.

Table 1:

Correlation of PET parameters and pathologic response

PET* Parameters	Median (Range) in %		p-value
	Pathologic Complete or Near-Complete Responders (n=11)	Pathologic Sub-Optimal Responders (n=16)
VRS†	65 (0 – 80)	33 (−8 – 67)	0.02
ΔSUVmax††	27 (−8 – 74)	23 (−31 – 69)	0.68
ΔSUVavg	34 (19 – 78)	34 (−38 – 79)	0.68
ΔTLG§	44 (32 – 91)	36 (−21 – 81)	0.15

Complete responders are defined as having 100% response. Near-complete responders are defined as having ≥ 95% response. Suboptimal responders are defined as having < 95% response.

^*PET, positron emission tomography;

^†VRS, visual response score;

^††SUV, standard uptake values;

^§TLG, total lesion glycolysis

Figure 2: ROC analysis for VRS.

Using a VRS cut-off of 50%, sensitivity = 94% and specificity = 55%.

(VRS, visual response score; ROC, receiver operating characteristic)

Table 2:

Description of patients by T-stage and N-stage

	ypT0N0	ypT0N1	ypT1N0	ypT2N0	ypT2N1	ypT2N2	ypT3N0	ypT3N1	Total
uT2N+	2	-	-	-	-	-	-	1	3
uT3N0	1	1	1	-	-	1	1	-	5
uT3N+	6			7	1		2	3	19
Total	9	1	1	7	1	1	3	4	27

Figure 3: Concordance Correlation Coefficient:

0.72 (indicates fair agreement)

Concordance between two PET readers regarding VRS

DISCUSSION

We have previously demonstrated that patients with locally advanced rectal cancer who achieve ≥ 95% pathologic response to preoperative CRT have improved long-term oncologic outcome.^[8] Therefore, early prediction of suboptimal pathologic response is of particular interest since it allows for immediate termination of ineffective therapy and pursuit of alternative treatment. We anticipate that early changes in therapy may improve extent of response in these patients, perhaps improving long-term outcome as well.

In our study, 11 patients (41%) had ≥ 95% pathologic response to preoperative CRT and were defined as pathologic complete or near-complete responders. Sixteen patients (59%) had < 95% pathologic response and were defined as suboptimal responders. By using the VRS, determined by comparing a baseline ¹⁸F-FDG-PET before CRT to another performed 8–14 days after commencement of CRT, suboptimal pathologic responders could be identified with a sensitivity and specificity of 94% and 55%, respectively.

In their 2001 study evaluating the efficacy of metabolic imaging for measuring response to preoperative treatment in patients with adenocarcinoma of the esophagogastric junction, Weber et al reported that a cut-off value of 35% decrease in FDG tumor uptake (SUV) between PET scans done prior to and 14 days after initiation of chemotherapy could predict pathologic response with a sensitivity and specificity of 93% and 95%, respectively.^[15] They concluded that response-guided treatment algorithms might be applied to clinical practice.^[18]

Although our results are based on only 27 cases, we do explore the novel concept of response-guided treatment of locally advanced rectal cancer. If we were to use the VRS cut-off point with the greatest accuracy (50%) to guide treatment (Figure 2), we would have changed treatment in 5 of 11 (45%) patients who were likely to have achieved a ≥ 95% pathologic response on their current treatment. However, this is balanced by the fact that 15 of 16 (94%) patients who were not likely to achieve a ≥ 95% pathologic response would have been identified early in treatment, in time to administer alternative therapy aimed at increasing their pathological response. This response-guided therapy is particularly appealing for colorectal cancer, given the recent development of cytotoxic and targeted therapies capable of yielding higher response rates.^[31–35]

Although we tested various PET parameters, including baseline SUV_avg, SUV_max and TLG, and changed these parameters from baseline to intermediate scans, only VRS showed significant association with pathologic response. This is consistent with our initial report, published in 2005.^[36] In 2006, however, in their study of 33 rectal cancer patients, Cascini et al reported that ΔSUV_avg was capable of successfully discriminating between pathologic response (which they defined as complete response, or residual cancer cells scattered through fibrosis) and non-response.^[37] The fact that different PET parameters appear to more effectively distinguish responders from non-responders may be explained, in part, by differences in study design and/or stage of disease as well as differences in preoperative CRT regimens. This is reflected in the higher rate of pathologic response reported by the Italian study (55%) compared with ours (41%). However, regardless of which individual or combined PET parameters may best predict pathologic response, it is clear that early metabolic assessment of rectal cancer response to preoperative CRT is a potentially promising management approach meriting further investigation. However, at this time, changes in PET parameters as determined in our study do not support a change in management of rectal cancer patients receiving preoperative chemoradiotherapy.

There are some limitations to this study. We reported VRS measurements based on consensus readings of ¹⁸F-FDG-PET scans, interpreted by two dedicated Nuclear Medicine physicians who were blinded to each patient’s clinical data. The level of agreement between both readers reached a concordance correlation coefficient of 0.72, which is low, but nevertheless indicates fair agreement beyond chance. Although we expected a higher level of agreement between the readers,^[38] the low concordance may be attributable to our relatively small sample size. Future studies with a larger sample size will need to demonstrate reproducibility of validated PET measurements in clinical practice before PET assessment of rectal cancer response to preoperative CRT might be considered useful.

CONCLUSIONS

In conclusion, early scanning using standard FDG-PET parameters did not accurately predict favorable from non-favorable responses. We have demonstrated that a visual response assessment based on ¹⁸F-FDG-PET scans performed prior to and 8–14 days after initiation of preoperative CRT, with a cut-off of 50%, predicts suboptimal pathologic response with a total accuracy of 78%. However, insofar as no viable alternative therapy currently exists, we do not believe that this level of accuracy is sufficient to prompt a change in therapy. VRS is a PET parameter that requires further validation.

SYNOPSIS.

We prospectively evaluated the accuracy of early FDG-PET in predicting patient response to preoperative chemoradiotherapy, and were unable to accurately discriminate favorable from unfavorable responders. Use of PET in assessing rectal cancer response to preoperative CRT remains experimental.