FDG-PET parameters as prognostic factor in esophageal cancer patients: a review.

BACKGROUND: (18)F-fluorodeoxyglucose positron emission tomography (FDG-PET) has been used extensively to explore whether FDG Uptake can be used to provide prognostic information for esophageal cancer patients. The aim of the present review is to evaluate the literature available to date concerning the potential prognostic value of FDG uptake in esophageal cancer patients, in terms of absolute pretreatment values and of decrease in FDG uptake during or after neoadjuvant therapy.
METHODS: A computer-aided search of the English language literature concerning esophageal cancer and standardized uptake values was performed. This search focused on clinical studies evaluating the prognostic value of FDG uptake as an absolute value or the decrease in FDG uptake and using overall mortality and/or disease-related mortality as an end point.
RESULTS: In total, 31 studies met the predefined criteria. Two main groups were identified based on the tested prognostic parameter: (1) FDG uptake and (2) decrease in FDG uptake. Most studies showed that pretreatment FDG uptake and postneoadjuvant treatment FDG uptake, as absolute values, are predictors for survival in univariate analysis. Moreover, early decrease in FDG uptake during neoadjuvant therapy is predictive for response and survival in most studies described. However, late decrease in FDG uptake after completion of neoadjuvant therapy was predictive for pathological response and survival in only 2 of 6 studies.
CONCLUSIONS: Measuring decrease in FDG uptake early during neoadjuvant therapy is most appealing, moreover because the observed range of values expressed as relative decrease to discriminate responding from nonresponding patients is very small. At present inter-institutional comparison of results is difficult because several different normalization factors for FDG uptake are in use. Therefore, more research focusing on standardization of protocols and inter-institutional differences should be performed, before a PET-guided algorithm can be universally advocated.

Esophageal cancer is an aggressive disease with early dissemination. Even after potentially curative surgery, long-term survival rates rarely exceed 35%.1,2 In order to improve this outcome, institutes apply neoadjuvant chemotherapy and/or radiotherapy; however, only patients who respond to this therapy benefit.3–7

Assessment of prognosis can influence patient management; a diagnostic test that provides pretreatment prognostic information will therefore have additional value. Moreover, prediction of tumor response early, during the neoadjuvant regimen, is of crucial importance. 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) is a noninvasive imaging technique that enables quantification of tumor activity on the basis of altered tissue glucose metabolism.8–10 Many studies have been published on the improvement of preoperative staging of esophageal cancer with FDG-PET by detecting distant metastases.11–13 FDG-PET also seems to be a valuable tool to monitor early response to neoadjuvant therapy.14–16 Evidence for reliable and useful response measurement in esophageal cancer patients is growing, while response measurement is already well established in, for example, non-small cell lung cancer and lymphoma.17–21

Recent literature suggests that FDG-PET at time of diagnosis might be useful for prognostication. The underlying idea is that the quantity of FDG activity in the tumor correlates with viable tumor cell number and thus with prognosis.22–26 The most commonly applied (semi-) quantification parameter in clinical PET is the standardized uptake value (SUV) of the primary tumor. SUV is determined by the ratio of activity in the region of interest (Bq/mL) over the decay-corrected activity of FDG injected into the patient (Bq/g).27,28

The present review evaluates the literature available to date concerning the potential prognostic value of FDG uptake in esophageal cancer patients, in terms of absolute pretreatment value and of decrease in FDG uptake during or after neoadjuvant therapy.

Literature Search

A review of the English language literature concerning esophageal cancer and standardized uptake values was performed. A computer-aided search was performed of the databases PubMed and Embase in January 2009. The terms “positron emission tomography,” “FDG-uptake,” “SUV,” and “esophageal cancer,” with restriction to the English language only, were used.29 All searches were performed using text word or medical subject heading (MeSH). Searches were focused on clinical studies evaluating the prognostic value of FDG uptake as an absolute value or the decrease in FDG uptake (during neoadjuvant therapy), possibly in combination with other factors, and using overall mortality and/or disease related mortality as an end point in esophageal cancer patients. Two researchers (J.M.T.O. and M.v.H.) read all abstracts and evaluated whether an abstract met the predefined criteria. After this selection, all publications were retrieved as full papers and re-evaluated for inclusion.

Results

In total, 31 studies met the predefined criteria.14,16,30–59 Two main groups were identified based on the tested prognostic parameter: (1) FDG uptake and (2) decrease in FDG uptake. In the first group, 15 studies described FDG uptake measured before any form of treatment was started (group 1A: Table 1), and 5 studies described FDG uptake measured after neoadjuvant treatment (group 1B: Table 2).30–44,52–56 In the second group, 6 studies described decrease in FDG uptake measured early during neoadjuvant therapy (group 2A: Table 3), and also 10 studies described decrease in FDG uptake measured after completion of neoadjuvant therapy (group 2B: Table 3).14,16,38,41,42,45–51,55,57–59 Also, 9 studies described the same cohorts of patients; however these were not excluded.31,32,35,36,41,42,47,49,51 Methodological aspects of included studies are described in Tables 4, 5, and 6.

Table 1

Characteristics of the 15 studies regarding pretreatment SUV and prognosis in esophageal cancer patients

Study/year of publication	Patients (n)	F/M	Age range (years)	AC/SCC/other	Stage diseasea	Treatment	SUV predictor of survival (univariate)	SUV independent predictor survival (multivariate)	Other independent predictive factors
Fukunaga/199833	48	5/43	44–76	ND	II–IV	Resection	Yes (OS)	–	–
Kato/200235	32	3/29	42–76	–/32/–	I–IV	Resection	Yes (OS)	–	–
Kato/200336	44	–	42–76	–/44/–	ND	Resection	Yes (OS)	–	–
Choi/200431	69	5/64	–	–/69/–	I–IV	Resection ± adjuvant CRT	Yes (OS)	No	cTNM, pTNM, PET-tumor length, PET + lnn
Hong/200534	47	4/43	36–78	41/6/–	II–III (cTNM)	Neoadjuvant CRT + resection	No	No	Number of PET abnormalities
Stahl/200540	40	–	–	40/–/–	II–IV (cTNM)	Resection ± neoadjuvant CT	No	–	–
van Westreenen/200543	40	16/24	48–79	28/12/–	I–IV (cTNM)	Resection/palliation	Yes (OS)	No	Resection
Cerfolio/200630	89	36/53	29–81	47/32/10	I–IV	Resection ± neoadjuvant CRT	Yes (OS)	Yes (OS)	TNM
Choi/200632	51	4/47	41–77	–/51/–	I–IV	Resection ± adjuvant CRT	Yes (OS)	No	pTNM, intratumoral MVD, PET + lnn, VEGF expression
Rizk/200639	50	6/44	–	50/–/–	I–IV	Resection	Yes (OS)	–	–
Westerterp/200844	26	3/23	48–79	26/–/–	I–IV	Resection ± Cox-2 inhibitor	Yes (DFS)	–	–
Omloo/200856	125	21/104	37–82	106/19/–	I–III	Resection	Yes (DFS)	No	EUS T-stage, tumor location, EUS N-stage, cTNM
Cheze-Le Rest/200853	47	5/42	41–89	11/36/–	I–IV	Resection ± neo/adjuvant CT ± RT	Yes (OS)	Yes (OS)	Treatment, number of PET abnormalities, PET + LNN, number of PET + LNN
Chatterton/200852	129	25/104	36–87	99/25/5	I–IV	Resection ± CT ± RT/palliation	No (DFS)	ND	Additional PET lesions
Makino/200855	38	7/31	50–76	–/38/–	I–IV	Neoadjuvant CT + resection	Yes (DFS)	No	PET + LNN, SUV decrease, pT, pN

SUV standardized uptake value, n number, F female, M male, AC adenocarcinoma, SCC squamous cell carcinoma, ND not described, OS overall survival, DFS disease-free survival, ± with or without, CRT chemoradiotherapy, CT chemotherapy, cTNM clinical TNM staging, pTNM pathological TNM staging, PET + lnn positive lymph nodes on PET, MVD microvessel density, VEGF vascular endothelial growth factor, Cox-2 cyclooxygenase-2

apTNM classification according to IUAC, unless stated otherwise

Table 2

Characteristics of the 5 studies regarding postneoadjuvant treatment SUV and prognosis in esophageal cancer patients

Study/year of publication	Patients (n)	F/M	Age range (years)	AC/SCC/other	Stage of diseasea	Treatment	SUV predictor of survival (univariate)	SUV independent predictor survival (multivariate)	Other independent predictive factors
Swisher/200441	83	9/74	34–79	73/10/–	0–IV	Neoadjuvant CRT ± induction CT + resection	Yes (OS)	No	–
Swisher/200442	103	12/91	34–79	90/13/–	II–IVa (cTNM)	Neoadjuvant CRT ± induction CT + resection	Yes (OS)	Yes (OS)	Esophageal wall thickness on CT (post-CRT)
Konski/200737	81	14/67	–	66/15/–	II–IVa (cTNM)	Definitive CRT/neoadjuvant CRT + resection	Yes, definitive CRT patients (OS)	No	No
Mamede/200738	25	3/22	ND	22/3/–	0–IVa	Neoadjuvant CRT + resection	Yes (DFS)	–	–
Higuchi/200854	50	9/41	44–77	–/50/–	III–IV	Neoadjuvant CT ± RT + resection	Yes (DFS)	ND	–

apTNM classification according to IUAC, unless stated otherwise

SUV standardized uptake value, n number, F female, M male, AC adenocarcinoma, SCC squamous cell carcinoma, CRT chemoradiotherapy, CT chemotherapy, OS overall survival, DFS disease-free survival, ± with or without, ND not described

Table 3

Characteristics of the studies regarding SUV decrease and prognosis early during neoadjuvant therapy (group 2A, 6 studies) and after completion of neoadjuvant therapy (group 2B, 10 studies)

Study/Year of publication	Patients (n)	F/M	Age range (years)	AC/SCC/other	Stage of diseasea	Treatment	Prevalence respondersb	SUV decrease predictor of response	SUV decrease predictor of survival	Absolute SUV available
Group 2A
Weber/200149	40	3/37	25–69	40/–/–	0–IV	Neoadjuvant CT + resection	31% (11/35)	Yes	Yes (OS + DFS)	Yes
Ott/200647	65	7/58	50–66	65/–/–	0–IV	Neoadjuvant CT + resection	18% (10/56)	Yes	Yes (OS)	Yes
Lordick/200714	119	8/111	ND	119/–/–	0–IV	Neoadjuvant CT + resection	69% (37/56)c	Yes	Yes (OS + DFS)	Yes
Wieder/200751	24	4/20	33–71	24/–/–	0–IV	Neoadjuvant CT + resection	33% (8/24)	Yes	Yes (OS)	Yes
Wieder/200450	38	11/27	46–73	–/38/–	0–IV	Neoadjuvant CRT + resection	58% (19/33)	Yes	Yes (OS)	Yes
Westerterp/200616	26	2/24	29–73	20/6/–	0–IVa	Neoadjuvant ThCRT + resection	42% (10/24)	Yes	Nod	Yes
Group 2B
Port/200748	62	10/52	36–76	51/11/–	0–IV	Neoadjuvant CT + resection	16% (10/62)	Yes	Yes (DFS)	No
Makino/200855	38	7/31	50–76	–/38/–	I–IV	Neoadjuvant CT + resection	59% (20/34)	–	Yes (DFS)	Yes
Downey/200345	39	5/34	36–76	26/13/–	0–III	Neoadjuvant CT ± RT + resection	24% (4/17)	–	No	No
Levine/200646	64	11/53	42–84	52/12/–	I–IV	Neoadjuvant CRT + resection	42% (20/48)	No	–	Yes
Mamede/200738	25	3/22	ND	22/3/–	0–IVa	Neoadjuvant CRT + resection	32% (8/25)	Yes	Yes (DFS)	Yes
Roedl/200958	49	10/39	ND	–/49/–	II–III	Neoadjuvant CRT + resection	45% (22/49)	–	No	No
Roedl/200857	51	5/46	ND	51/–/–	I–IVa	Neoadjuvant CRT + resection	41% (21/51)	Yes	Yes (DFS)	No
Schmidt/200959	55	12/43	34–74	31/24/–	III–IVa	Neoadjuvant CRT + resection	38% (21/55)	No	No	Yes
Swisher/200441	83	9/74	34–79	73/10/–	0–IV	Neoadjuvant CRT ± induction CT + resection	54% (43/79)	–	–	Yes
Swisher/200442	103	12/91	34–79	90/13/–	II–IVa (cTNM)	Neoadjuvant CRT ± induction CT + resection	56% (58/103)	No	No	Yes

SUV standardized uptake value, n number, F female, M male, AC adenocarcinoma, SCC squamous cell carcinoma, CRT chemoradiotherapy, CT chemotherapy, ThCRT thermochemoradiation therapy, OS overall survival, DFS disease-free survival, ND not described

apTNM classification according to IUAC

bHistopathology not available in some patients; no surgery due to disease progression

cResponse rate in patients classified as metabolic responders after 2 weeks of CT

dAt a median follow-up of only 9 months all responders were still alive

Table 4

Methodological aspects of FDG uptake used as absolute value to predict prognosis in esophageal cancer patients

Study/Year of publication	Single/Multicenter	Scanner	Reconstruction methods	ROI methods	Injected dose FDG (MBq)	Time between injection and scan	Quantification method	SUV max or SUV mean iso 50%/70%	Corrected for	Plasma glucose measurements	Absolute values (SUV, range)	Cutoff values
Fukunaga/199833	Single	HEAD-TOME III (Shimazu Works, Kyoto, Japan)	Ramp-filter + Butterworth filter, 10.5 mm FWHM	Site of maximum accumulation (9 pixels: 9 × 9 mm2)	148	60 min	SUV	SUV max	BW	Noa	1.51–16.13	7.0
Kato/2002, 200335,36	Single	SET 2400W (Shimadzu Corporation, Kyoto, Japan)	OSEM, 4.2 mm FWHM	Manually drawn 1 cm in dimension at site of tumor	275–370	40 min	SUV	SUV max	BW	No	1.43–9.0	3.0
Choi/2004, 200631,32	Single	Advance PET scanner (General Electric Medical Systems, Milwaukee, WI)	FBP, Hanning-filter, 8.0 mm	ND	370	45 min	SUV	SUV max	BW	No	–	6.3, 13.7
Hong/200534	Single	ND	ND	ND	ND	ND	SUVb	Peak SUV, SUV primary and total SUV	ND	No	–	4.0
Stahl/200540	Single	ECAT EXACT (Siemens, Knoxville, TN)	OSEM 8 iterations/4 subsets, 3D Gaussian filter 4 mm FWHM	Manually placed circular ROI of 1.5 cm on tumor site maximal FDG accumulation	400	90 min	SUV	SUV max	BW	Yes	–	10.5
van Westreenen/200543	Single	ECAT EXACT HR+ (Siemens/CTI, Knoxville, TN)	OSEM, filter ND	3D ROI selected semi-automatically	130–690	90 min	SUV	SUV max and SUV mean iso 70%	BW	No	1.8–19.2	6.7
Cerfolio/200630	Single	ECAT EXACT (CTI, Knoxville, TN)/integrated PET-CT (Discovery LS, General Electric, Milwaukee, WI)	ND	Manually drawn ROI around tumor	555	60 min	SUV	SUV max	BW	No	–	6.6
Rizk/200639	Single	Advance PET scanner (General Electric Medical Systems, Milwaukee, WI)/CTI Biograph (CTI, Knoxville, TN)	ND	ROI analysis tools delivered with scanner	370–555	ND	SUV	SUV max	BW	No	1.9–19.1	4.5
Konski/200737	Single	Integrated PET-CT (Discovery LS, General Electric, Waukesha, WI)	2D, OSEM 2 iterations/28 subsets, Gaussian filter 10 mm FWHM	ND	370–740	90–120 min	SUVb	SUV max	BW	Yes	–	–
Westerterp/200844	Single	ECAT EXACT HR+ (Siemens/CTI, Knoxville, TN)	2D, OSEM 2 iterations/16 subsets, Gaussian filter 5 mm FWHM	VOI generated by 3D region-growing algorithm with in-home developed software	350–597	90 min	SUV	SUV max and SUV mean iso 50%	BSA, glucose	Yes	0.03–0.63	0.26
Omloo/200856	Multi	ECAT EXACT HR+ (Siemens/CTI, Knoxville, TN)	2D, OSEM 2 iterations/16 subsets, Gaussian filter 5 mm FWHM	VOI generated by 3D region-growing algorithm with in-home developed software	130–810	90 min	SUV	SUV max and SUV mean iso 50%	BSA, glucose	Yes	0.13–0.45 (IQR)	0.27
Cheze-Le Rest/200853	Single	Allegro-dedicated PET scanner (Philips Medical System, Cleveland, OH)	3D RAMLA reconstruction protocol	ROI analysis highest uptake	5 MBq/kg	60 min	SUV	SUV max	BW	Yes	9.3 ± 3.9 (mean, 1SD)	9
Chatterton/200852	Single	ND	ND	ND	120–400	45 min	SUV	SUV max	BW	No	–	8.2
Higuchi/200854	Single	PET scanner HEADTOME/set 2400W (Shimadzu Co, Kyoto, Japan)	ND	ROI selected semiautomatically	370	60 min	SUV	SUV max	BW	Yes	–	2.5

FDG fluorodeoxyglucose, ND not described, mm millimeters, FBP filtered backprojection, FWHM full width half maximum, OSEM ordered subset expectation maximization, 2D two-dimensional, ROI region of interest, cm centimeters, VOI volume of interest, SUV standardized uptake value, BW body weight, min minutes, hrs hours, BSA body surface area

aChanges of radioactivity in plasma and tumor (rate constants, k1–k4) were calculated

bSUV was used to quantify FDG uptake; however, SUV methods were not described

Table 5

Methodological aspects of decrease in FDG uptake during early response monitoring to predict prognosis in esophageal cancer patients

Study/Year of publication	Single/Multicenter	Scanner	Reconstruction methods	ROI methods	Injected dose FDG (MBq)	Time between injection and scan	Quantification method	SUV max or SUV mean iso 50%/70%	Corrected for	Plasma glucose measurements	Absolute values (SUV, range)	Cutoff value responding vs nonresponding
Weber/200149 a	Single	ND	FBP, Hanning filter 0.4, 6–8 mm FWHM	Manually placed circular ROI of 1.5 cm on tumor site maximal FDG accumulation	250–370	40 min	SUV	SUV max	BSA	Yes	5.0–50.3	−35%
Wieder/200450	Single	ECAT EXACT (Siemens/CTI, Knoxville, TN)	OSEM 8 iterations/4 subsets, 3D Gaussian filter 4 mm FWHM	Manually placed circular ROI of 1.5 cm on tumor site maximal FDG accumulation	300–400	60 min	SUV	SUV max	BW	Yes	0.9–15.4	−30%
Westerterp/200616	Single	ECAT EXACT HR + (Siemens/CTI, Knoxville, TN)	2D, OSEM 2 iterations/16 subsets, Gaussian filter 5 mm FWHM	VOI generated by 3D region growing algorithm with in-home developed software	250–370	90 min	SUV	SUV mean iso 50%	BSA, glucose	Yes	0.1–0.5	−31%
Ott/200647 a	Single	ND	OSEM 8 iterations/4 subsets, 3D Gaussian filter 4 mm FWHM	Manually placed circular ROI of 1.5 cm on tumor site maximal FDG accumulation	250–370	40 min	SUV	SUV max	BW	Yes	–	−35%
Wieder/200751	Single	ECAT EXACT (Siemens/CTI, Knoxville, TN)	OSEM 8 iterations/4 subsets, 3D Gaussian filter 4 mm FWHM	Manually placed circular ROI of 1.5 cm on tumor site maximal FDG accumulation	300–400	40 min	SUV	SUV max	BW	Yes	–	−33%
Lordick/200714 a	Single	ECAT EXACT full ring (Siemens/CTI, Knoxville, TN)	3D, OSEM 8 iterations/4 subsets, FBP Hanning filter 0.4, 6–8 mm FWHM	Manually placed circular ROI of 1.5 cm on tumor site maximal FDG accumulation	300–400	40 min	SUV	SUV max	BSA	Yes	–	−35%

aSome patients were included in all 3 studies

FDG fluorodeoxyglucose, ND not described, mm millimeters, FBP filtered backprojection, FWHM full width half maximum, OSEM ordered subset expectation maximization, 2D two-dimensional, ROI region of interest, cm centimeters, VOI volume of interest, SUV standardized uptake value, BSA body surface area, min minutes, hrs hours, BW body weight

Table 6

Methodological aspects of decrease in FDG uptake after completion of neoadjuvant therapy to predict prognosis in esophageal cancer patients

Study/year of publication	Single/multicenter	Scanner	Reconstruction methods	ROI methods	Time between injection and scan	Injected dose FDG (MBq)	Quantification method	SUV max or SUV mean iso 50%/70%	Corrected for	Plasma glucose measurements	Absolute values (SUV, range)	Cutoff value responding vs nonresponding
Downey/200345	Single	Advance PET scanner (General Electric Medical Systems, Milwaukee, WI)	FBP, ND	ROI analysis tools delivered with scanner	ND	>370	SUV	SUV max	ND	No	–	−60%
Swisher/200441,42	Single	ECAT EXACT HR+ (Siemens/CTI, Knoxville, TN)	OSEM 2 iterations/8 subsets, Gaussian filter 4.5 mm FWHM	Manually placed ROI on tumor site with FDG accumulation	45/60 min	555–740	SUV	SUV max	BW	No	–	4.0 (SUV)
Levine/200646	Single	Advance PET scanner (General Electric Medical Systems, Milwaukee, WI)	ND	ROI analysis tools delivered with scanner	60 min	555–740	SUV	SUV max	LBM	Yes	0–36.6	>−10.0 (SUV)
Port/200748	Single	Advance PET scanner (General Electric Medical Systems, Milwaukee, WI)	ND	ND	45–60 min	370–555	SUV	SUV max	ND	No	–	−50%
Mamede/200738	Single	Integrated PET-CT (Discovery LS, General Electric, Milwaukee, WI)	OSEM 2 iterations/30 subsets	Manually placed circular ROI of 1.5 cm on tumor site maximal FDG accumulation	±80 min	PET 1: 813 ± 144 PET 2: 720 ± 91	SUV	SUV max	BW	Yes	–	−32%
Roedl/200858	Single	Biograph 16 integrated PET/CT scanner (Siemens, Erlangen, Germany)	ND	Delineated automatically including pixels equal/greater to SUV 2.5	60 min	555	SUV	SUV max + SUV mean	BW	No	–	55% (diameter SUV index)
Makino/200855	Single	HEADTOME/SET 2400W (Shimadzu, Kyoto, Japan)	Iterative median root + reconstruction algorithm, filter 3.7 mm FWHM	ROI of 10 pixels on tumor site maximal FDG accumulation	60 ***min	370	SUV	SUV max	BW	Yes	11.12 ± 4.32 (mean ± SEM)	−70%
Roeld/200857	Single	Biograph 16 integrated PET/CT scanner (Siemens, Erlangen, Germany)	ND	Delineated automatically including pixels equal/greater to SUV 2.5	60 min	555	SUV	SUV max + SUV mean	BW	No	–	−63% (PET-CT volume)
Schmidt/200859	Single	ECAT EXACT 47 scanner (Siemens Medical Systems, Siemens CTI, Knoxville, TN)	OSEM 2/iterations, 8 subsets, Gaussian filter 6 mm FWHM	Circular 10 pixel standard region + spherical ROI in maximal FDG accumulation	60 min	370	SUV	SUV max + SUV mean	BW	Yes	1.8–19.4	SCC, −70% AC, −22%

FDG fluorodeoxyglucose, ND not described, mm millimeters, FBP filtered backprojection, FWHM full width half maximum, OSEM ordered subset expectation maximization, 2D two-dimensional, ROI region of interest, cm centimeters, VOI volume of interest, SUV standardized uptake value, ND not described, hrs hours, BW body weight, min minutes, LBM lean body mass, SCC squamous cell carcinoma, AC adenocarcinoma

FDG Uptake as Prognostic Factor (Group 1)

Group 1A: Pretreatment FDG Uptake and Prognosis (Table 1)

In 1998 Fukunaga et al. found in 48 patients that even though clinicopathological findings did not correlate with FDG uptake, patients with a high SUV had a poorer prognosis compared with those with low FDG uptake (55% 2-year disease-free survival vs 30%).33 This study is limited by the lack of multivariate analysis. In 2002 Kato et al. showed that FDG uptake was associated with depth of tumor invasion, presence of lymph node metastases, and lymphatic vessel invasion in 32 patients.35 The 2-year survival rate in patients with high FDG uptake (48%) was lower than in patients with low FDG uptake (91%). It would have been helpful if the authors had provided 95% confidence intervals for these survival rates. In another publication on partly the same cohort, a significant correlation was found between FDG uptake and Glut-1 expression; low Glut-1 expression and low FDG uptake appeared to carry a better prognosis: these patients showed 100% 2-year survival (n = 15).36 Multivariate analysis was unfortunately not performed.

Choi et al. showed in a multivariate analysis that only PET + lnn was an independent prognostic factor for disease-free survival.31 In multivariate analysis for overall survival only cTNM, pTNM, PET tumor length, and PET + lnn were independent predictive factors. The large proportion of patients with squamous cell carcinomas included in this study limits the use of these results in western populations. In another publication on partly the same cohort multivariate analysis showed pTNM, PET + lnn, VEGF expression, and intratumoral microvessel density (MVD) to be independent predictors for overall survival. A total of 7 variables were included in the multivariate regression model, well exceeding the generally acceptable number of 1 variable per every 10 events and thus increasing the risk of coincidental findings.32

Hong et al. showed in 47 patients with locoregional esophageal cancer that the number of PET abnormalities (NPA) correlates with overall and disease-free survival in univariate and multivariate analysis, while FDG uptake did not.34 Only half of the patients underwent esophagectomy (no explanation provided). Clinical TNM stage was not included in this analysis to assess independent value of NPA. Stahl et al. showed in a retrospectively analyzed cohort of 40 patients with esophageal cancer that FDG uptake in the primary tumor did not correlate with overall survival.40 The authors suggest that the reason for this might be because they only included adenocarcinomas.

Van Westreenen et al. investigated the relation between FDG uptake and the stage of disease and whether FDG uptake could be used to predict resectability and survival in 40 retrospectively collected patients with any stage of disease.43 Patients with high FDG uptake had a worse mean survival rate compared with patients with low FDG uptake (9 months compared with 20 months; P = .02). Patients eligible for resection showed a significantly lower FDG uptake compared with those with irresectable disease.

Cerfolio and Bryant showed in a multivariate analysis that patients with high FDG uptake were more likely to have poorly differentiated tumors and advanced stage using a retrospective cohort of 89 patients.30 Remarkably, FDG uptake correlated better with survival than pathological TNM stage. The 4-year survival of patients with low FDG uptake was 89% and only 31% in patients with high FDG uptake. It was, however, stated that many different pathologists with unspecified experience were used for staging the resection specimens.

Rizk et al. found that 3-year survival was 95% for patients with low FDG uptake and 57% for patients with high FDG uptake, in a retrospective analysis of 50 patients with resectable adenocarcinoma of the distal esophagus.39 The survival advantage for patients with low FDG uptake was even seen in a subset of patients with clinically and pathologically early-stage disease. This finding is quite remarkable considering the range of survival in this group of patients compared with a group of patients with all stages of disease.

Westerterp et al. investigated biological parameters to predict in which patients FDG-PET could be of prognostic value, in 26 patients.44 No association was found between FDG uptake and angiogenic markers, hexokinase isoforms, Ki-67 antigen expression, cleaved caspase-3, cell density, differentiation grade, CD68, mucus, or necrosis. Glut-1 expression showed a significant correlation with FDG uptake. They concluded that Glut-1 may be used to select esophageal cancer patients in whom FDG-PET is of diagnostic value. Even in the subgroup of patients who underwent a microscopically radical resection a strong association was found between SUV and survival (P = .001).

In one of the largest available prospective studies, Omloo et al. assessed the prognostic importance of SUV and EUS parameters.56 In 125 patients who underwent esophagectomy without neoadjuvant therapy SUV, tumor location, EUS T-stage, EUS N-stage, and clinical stage proved to be of prognostic significance in univariate analysis. In multivariate analysis, however, EUS T-stage appeared to be the only independent predictor for survival.

Cheze-Le Rest et al. investigates a total of 52 patients with all stages of disease; performance of potentially curative surgery, SUVmax >9 and 2 or more PET abnormalities were significant prognostic predictors.53 In multivariate analysis, only SUVmax >9 and the presence of FDG-positive lymph nodes were found as independent predictors of poor outcome. Notably, 2 of 3 PET-derived parameters were almost identical: presence of >1 FDG-PET positive node and presence of >2 FDG-PET positive nodes. In the largest available study Chatterton et al. aimed to determine the impact of PET on clinical management and prognosis in 129 potentially curable patients.52 Significant changes in management were observed in 38% of patients, primarily as a result of the identification of additional sites.

Makino et al. found SUVmax <12 and the number of positive lymph nodes (PET + LNN) on PET before therapy to be of prognostic significance in a retrospective cohort of 38 patients with positive lymph nodes scheduled to undergo neoadjuvant chemotherapy.55 Unfortunately only 38 of 63 patients who met the inclusion criteria were included.

In summary, most studies (12 of 15) showed that pretreatment FDG uptake is a predictor for survival in univariate analysis, whereas only 2 studies showed FDG uptake to be a predictor of survival in multivariate analysis.30–33,35,36,39,43,44,53,55,56 More importantly, neither of the 2 largest prospective trials could prove the prognostic significance of FDG-PET.52,56

Group 1B: Residual Postneoadjuvant Treatment FDG Uptake and Prognosis (Table 2)

In a prospective trial, Swisher et al. reported postneoadjuvant treatment FDG-PET uptake to be able to predict response, but failed to accurately rule out microscopic residual tumor (R1 resection) in 18% of a total of 83 patients.41 Swisher et al. evaluated a similar cohort of patients to assess the utility of PET, endoscopic ultrasonography (EUS), and CT to predict pathologic response and survival.42 FDG uptake was most accurate to predict long-term survival after neoadjuvant therapy. As before, they concluded that FDG uptake cannot rule out residual disease and that esophagectomy should remain part of the therapy. Because many of the patients in this study also seem included in the previously described study by Swisher et al., these reports should not be regarded as 2 separate studies.41

Konski et al. found a correlation between the depth of tumor invasion (determined by endoscopic ultrasonography) and the baseline FDG uptake in 81 patients undergoing definitive or preoperative chemoradiotherapy.37 Only posttreatment FDG uptake predicted disease-free survival in the definitive chemoradiotherapy group. The authors state to be cautious when using posttreatment FDG uptake to determine the necessity of surgical resection, as in this group of patients no correlation between FDG uptake and disease-free survival was found. It remains unclear which variables were used in multivariate analysis, complicating data interpretation.

In a relatively small study Mamede et al. showed that FDG uptake measured before treatment correlated with clinical T stage, advanced clinical stage, tumor length, and tumor volume as determined on PET.38 FDG uptake measured after treatment was the best predictor of disease progression. The authors conclude that FDG uptake should have a definite role in the evaluation of response to therapy and in the prediction of progression-free survival, which seems rather progressive considering the number of included patients (n = 25).

Higuchi et al. showed low FDG uptake after neoadjuvant treatment to be predictive for long-term survival (P = .0071); SUV was measured in 29 of 50 patients who were included.54 Unfortunately, multivariate analysis including histopathological response was not performed.

In summary, all 5 studies showed that FDG uptake after neoadjuvant therapy was predictive for survival in univariate analysis; however, in multivariate analysis only 1 study showed FDG uptake to be independently predictive for survival.37,38,41,42,54

Decrease in FDG Uptake as Prognostic Factor (Group 2)

Group 2A: Decrease in FDG Uptake Early During Neoadjuvant Treatment and Prognosis (Table 3)

In 2001 Weber et al. evaluated in a small but well-performed study whether reduction of FDG uptake can predict response 14 days after start of neoadjuvant chemotherapy.49 A significant difference in reduction of FDG uptake was found between responding (−54%) and nonresponding patients (−15%). Applying the optimal ROC-derived cutoff value of 35% reduction as criterion for metabolic response, FDG-PET predicted histopathological response with a sensitivity of 93% (14 of 15 patients) and a specificity of 95% (21 of 22). Patients without metabolic response were characterized by significantly shorter 2-year overall survival (37% vs 60%, P = .04).

This same group of investigators validated the previous findings using this definition of metabolic response, using 65 patients.47 Metabolically responding patients showed a high histopathologic response rate (44%) with a 3-year survival rate of 70%. Metabolically nonresponding patients showed a histopathologic response rate of only 5%, and a 3-year survival rate of 35% (P = .01). The authors concluded that this study provides the basis for clinical trials in which preoperative treatment is discontinued for patients without metabolic response early in the course of therapy.

To assess the feasibility of a PET-response-guided treatment algorithm, the same group of investigators conducted a prospective single-center study, including 119 patients all of whom underwent 2 weeks of neoadjuvant chemotherapy and subsequent evaluation.14 After 2 weeks, metabolic responders (FDG uptake decrease >35%) continued to receive neoadjuvant chemotherapy for 12 more weeks; nonresponders discontinued neoadjuvant treatment and proceeded to immediate surgery. In addition, 58% of the metabolic responders also appeared to be histopathological responders. Median disease-free survival in metabolic responders was 30 months compared with 14 months in metabolic nonresponders. These results could at least partly be explained by the fact that metabolic responders underwent a total of 14 weeks of chemotherapy, whereas nonresponders only had 2 weeks of chemotherapy.

In another study from this same group of investigators, FDG-PET was performed before initiation of chemotherapy, 14 days after the start and preoperatively in 24 patients.51 Changes in FDG uptake at both time points were significantly correlated with histopathologic response, and reduction in FDG uptake early in the course of therapy was also significantly correlated with survival (P = .03).

In 2004 Wieder et al. analyzed 38 patients with squamous cell carcinomas treated with neoadjuvant chemoradiotherapy and subsequent esophagectomy.50 Histopathological responders showed a decrease of 44% in FDG uptake after 2 weeks of therapy, compared with 21% in histopathological nonresponders (P = .06). Metabolic changes were significantly correlated with survival (P = .01).

In 2006 Westerterp et al. performed FDG-PET before start and after 14 days of neoadjuvant thermochemoradiotherapy.16 In histopathological responders the median decrease in FDG uptake was 44%, compared with 15% in nonresponders. At a cutoff value of 31% decrease in FDG uptake compared with baseline, sensitivity to detect response was 75% with a corresponding specificity of also 75%.

In summary, all 6 of the aforementioned studies showed that early decrease in FDG uptake is predictive for pathological response. All but 1 study showed decrease in FDG uptake also to be predictive for survival.16 Unfortunately, 5 of 6 of these studies were performed in 1 single institute, underlining the need for new multicenter studies to confirm these findings.

Group 2B: Decrease in FDG Uptake Postneoadjuvant Treatment and Prognosis (Table 3)

Port et al. retrospectively reviewed the ability of FDG-PET to predict clinical and pathological response to preoperative chemotherapy in 62 patients.48 Almost 60% of the patients showed ≥ 50% decrease in FDG uptake, showing a better survival compared with metabolically nonresponding patients (36 vs 18 months, P = .03). Multivariate analysis showed metabolic response to be the only significant predictor for disease-free survival. Including 5 variables in a multivariate model with roughly 60 patients and 30 events is, however, a stretch.

Makino et al. found that patients with a decrease in SUV above the cutoff value of 70% showed significantly better survival.55 Decrease in uptake in the primary tumor as well as in lymph nodes were associated with survival.

In 2003 Downey et al. found that stratification below or above 60% decrease in FDG uptake leads to a 2-year survival of 38% in metabolic nonresponders compared with 67% for metabolic responders (P = .06).45 No details were provided as to why only 39 of a total of 184 patients were included in this study.

In 2006 Levine et al. evaluated a total of 64 patients who underwent PET before the initiation of therapy and 4–6 weeks after completion of therapy.46 A decrease in absolute FDG uptake was predictive of histopathological response (P = .05), not for survival.

The study of Mamede et al. found a 32% decrease in FDG uptake to be the best cutoff value for histopathological response with 75% sensitivity and 63% specificity and for disease-free survival.38

Roedl et al. found the highest accuracy to predict response and survival using the decrease of the diameter-SUV index, a decrease of 55% or more identified pathologic responders with a sensitivity of 91% and a specificity of 93%.58 Metabolic responders had a mean disease-free survival of 32 months, nonresponders 16 months (P = .001).

In another study of Roedl et al., 51 patients with adenocarcinoma were studied.57 Decrease in tumor volume appeared to be a better predictor for response and survival compared with decrease in SUV. The highest accuracy was achieved using the total lesion glycolysis (calculated by multiplying the tumor volume using the mean SUV of the volume) to identify treatment responders.

Schmidt et al. found neither baseline nor preoperative nor SUV reduction to correlate significantly with response or survival in 55 patients treated with neoadjuvant chemoradiotherapy.59

In summary, decrease in FDG uptake after completion of neoadjuvant therapy was predictive for response and survival in only 4 of 10 studies.38,48,55,57 Remarkably, these studies included fewer patients and showed lower percentages of responding patients compared with the other 6 studies. Despite some positive findings, none of these studies suggests that these posttreatment prediction models should have any therapeutic consequences.

Discussion and Conclusion

Most studies showed that pretreatment FDG uptake and postneoadjuvant treatment FDG uptake as absolute values are predictors for survival in univariate analysis. Moreover, early decrease in FDG uptake during neoadjuvant therapy is predictive for response and survival in most studies described. However, late decrease in FDG uptake after completion of neoadjuvant therapy was predictive for response and survival in only 2 of 6 studies. A major disadvantage is that some studies included patients with a wide range of disease (adenocarcinomas and squamous cell carcinomas, stage I through IV) and studies used different neoadjuvant treatment regimens. Especially those studies that describe patients receiving radiotherapy, it is known FDG uptake in these patients remains higher compared with patients receiving only chemotherapy. Most importantly, all institutes used different scanners with different protocols and used different reconstruction methods, and these heterogeneous data made pooling of results impossible.

Many prognostic factors, determined pretreatment and/or posttreatment, for example, TNM stage, histopathology results, and PET-derived parameters (including SUV, metabolic tumor volume, and total lesion glucolysis) are used to predict survival in esophageal cancer patients.60 In clinical practice, these factors are communicated with the patient to choose the most appropriate therapy. However, before a PET-guided treatment algorithm can be reliably implemented, more research focusing on standardization of protocols and inter-institutional technical differences should be performed in larger patient cohorts.

To date, it is difficult to compare results from different institutes and more importantly, published cutoff values are method specific and often institute specific, especially since they are also affected by acquisition protocol, reconstruction algorithm, and region of interest definition.61,62 Most importantly, to overcome these problems large multicenter prospective trials are necessary.

In conclusion, FDG-PET seems to be useful for prognostication and (neo)adjuvant treatment response assessment in esophageal cancer. However, more attention has to be paid in standardization of FDG-PET acquisition and reconstruction.