The higher the decrease in the standardized uptake value of positron emission tomography after chemoradiation, the better the survival of patients with gastroesophageal adenocarcinoma

Abstract

BACKGROUND:

Postchemoradiation percentage decrease in standardized uptake value (SUV) of positron emission tomography (PET) from baseline correlates with overall survival (OS) and pathologic response. Analyses of dichotomized data are commonly reported. The authors analyzed percentage SUV decrease as both dichotomized and continuous variables.

METHODS:

The authors assessed 151 consecutive patients with gastroesophageal adenocarcinoma who had chemoradiation and surgery. Baseline and postchemoradiation PET/computed tomography imaging was performed. The log-rank test and Cox proportional hazards models were used to associate percentage SUV changes and OS, and logistic regression models were used to detect the association between percentage SUV changes and pathologic response.

RESULTS:

A >52% SUV decrease (dichotomized analysis) was associated with a longer OS (log-rank test, P = .023). The univariate Cox proportional hazards model indicated that greater percentage SUV decrease (as a continuous variable) was associated with a lower risk of death (hazard ratio [HR], 0.99; P = .01). Pathologic response (≤50% residual cancer) was associated with longer OS (P = .003). Patients with chemoradiation resistance (>50% residual cancer) tended to have a higher risk of death than those with chemoradiation sensitivity (0-50% residual cancer; HR, 2.12; P = .099). In the multivariate model, the percentage SUV decrease (as a continuous variable) was the only prognosticator of OS (P = .01). The percentage SUV decrease was nonsignificantly associated with pathologic complete response (univariate odds ratio [OR], 1.01; P = .06 and multivariate OR, 1.03; P = .07).

CONCLUSIONS:

The greater the decline in SUV after chemoradiation, the longer is the OS of gastroesophageal adenocarcinoma patients. The percentage SUV decrease as a continuous variable is a better prognosticator of OS than its dichotomized assessments. Cancer 2009. © 2009 American Cancer Society.

Carcinoma of the esophagus is a significant global health problem, with >600,000 new patients expected each year.1 In the West, the incidence of esophageal adenocarcinoma has been increasing on a consistent basis for more than 30 years2 (www.cancer.org, accessed July 7, 2008). Patients are often diagnosed in an advanced stage, and the mortality continues to climb.2, 3 The most common stages for patients with localized esophageal adenocarcinoma are II and III. Patients with localized esophageal cancer (>T1b) are often treated with preoperative chemoradiation in Western universities and cancer centers, although the published results from randomized trials have been equivocal.4-7 One of the major frustrations with the trimodality therapy of esophageal cancer is related to considerable toxicities and complications,8, 9 and therapy can have serious lifestyle-altering consequences.

The other confounding factor is the unpredictable patient outcomes even in patients with the same clinical stage when they are treated similarly. The differing clinical outcome with the same therapy is because of the inherent genetic makeup of individual esophageal cancers (like other cancers) dictated by genetic and epigenetic alternations.10-12 A patient's own genetic makeup most likely contributes to the response and tolerance to therapy.13, 14 To optimize therapy for patients with gastroesophageal cancer, tools are needed to individualize therapy. Currently, clinical parameters do not predict outcome before the start of therapy.15-18 However, the initial results and subsequent changes in positron emission tomography (PET) show promise.19-25

Many investigators have demonstrated that decrease in metabolic activity as measured by the standardized uptake value (SUV) on PET/computed tomography (CT) as a result of therapy is of prognostic value.25-28 In addition, PET/CT images obtained during therapy have been studied in a limited number of patients and seem to provide useful information regarding the response to therapy and prognosis of patients.21, 29 However, the PET/CT data in gastroesophageal adenocarcinoma have not been validated and need further investigation.

Investigators have often reported on relatively small numbers of patients and have emphasized fixed cutpoints (dichotomized results) in percentage SUV decrease21, 25, 27, 29 or have used a specific SUV value as a cutpoint.25, 28 Here we report on the treatment-induced percentage SUV decrease both as a dichotomized (as suggested by the literature) and as a continuous variable in 1 of the largest series of patients with gastroesophageal adenocarcinoma undergoing chemoradiation and surgery, for the prediction of overall survival (OS) and pathologic response.

MATERIALS AND METHODS

Patient Selection

Patients with potentially resectable, biopsy-proven adenocarcinoma locally confined to either the thoracic esophagus or the gastroesophageal junction, evaluated at The University of Texas M. D. Anderson Cancer Center from 2002 to early 2006, were eligible. Resectability and operability were determined by patient evaluation by a multidisciplinary team including medical oncologists, gastroenterologists, radiation oncologists, radiologists, and thoracic oncologic surgeons. Patients with type 1 or type 2 (by the Siewert classification30) gastroesophageal cancer and M1a disease were included, whereas T1N0 and M1b (unresectable) cancer were excluded, because these patients are not treated with trimodality therapy at our institution.

Pretreatment investigations included a complete blood count, measurement of serum electrolytes, chest radiograph, CT scan of the chest and abdomen, barium swallow radiography, upper gastroesophageal endoscopy with endoscopic ultrasound, and a PET scan. All patients had another PET scan done 5 to 6 weeks after the completion of chemoradiation and just before surgery.

Study Design

The objectives of this analysis were 1) to correlate percentage SUV changes occurring from the initial PET/CT scan to postchemoradiation PET/CT with pathologic response and complete pathologic response and 2) to correlate percentage SUV changes occurring from the initial PET/CT scan to postchemoradiation PET/CT with OS.

Chemoradiotherapy

All patients received 5-fluorouracil concurrently with radiation. The additional drug was either a platinum compound or taxane. Some patients received an agent from all 3 classes of cytotoxic agents. The total radiation dose delivered was either 45 grays (Gy) in 25 fractions or 50.4 Gy in 28 fractions, at 1.8 Gy per fraction delivered once daily, 5 days per week.

Surgery

Patients were evaluated before surgery using barium swallow radiography, CT scan of the chest and abdomen, esophagoscopy, and PET/CT. All patients underwent surgery. Surgical procedures performed included 3-field esophagectomy, transhiatal esophagectomy, transthoracic esophagectomy, or minimally invasive esophagectomy. The resected surgical specimen from each patient was examined to determine the degree of pathologic response based on the criteria described previously.18, 31

Information on OS was obtained for each patient. Data were collected from the hospital records, patient follow-up data, death records in the public databases, and the M. D. Anderson tumor registry.

Pathologic response criteria

We used the criteria for rating the surgical specimen previously described by our group18 and validated by other institutions.31 Resistance to chemoradiation was designated when >50% residual cancer was noted in the surgical specimen, and response to chemoradiation was designated when ≤50% residual cancer was noted in the surgical specimen. Pathologic complete response was designated when no cancer cells were noted in the resected specimen.

PET/CT Scans

Fluorodeoxyglucose PET/CT scans were performed on a dedicated PET/CT system (Discovery ST, STe, or RX, General Electric Medical Systems, Milwaukee, Wis). Scan coverage was generally from the orbits to the proximal thighs, although coverage was altered to answer specific clinical questions and also sometimes included the entire head and/or the lower extremities. Scans were acquired 60 to 90 minutes after intravenous administration of [18F]fluorodeoxyglucose with a dose range of 15 to 20 mCi (555-740 MBq). PET studies were acquired in either 2-dimensional (2D) or 3D acquisition mode at 3 to 5 minutes per bed position (depending on the patient body mass index). Images were reconstructed using ordered-subset expectation maximum with a 128 × 128 matrix size and a field of view of 70 cm (5.47-mm pixels). CT was acquired without oral or intravenous contrast material, and was used for attenuation correction of the PET scan. The CT acquisition parameters were 120 kilovolt peaks, 300 milliAmperes, 0.5 seconds rotation, with a pitch of 1.375. The CT images were reconstructed using a 3.75-mm slice thickness with a slice interval of 3.27 mm to match the PET data. PET/CT images were reviewed on a workstation (Advantage Workstation, General Electric Medical Systems, Milwaukee, Wis). PET, CT, and PET/CT fusion datasets were reviewed in multiple imaging planes.

SUV Calculations

The maximum SUV (SUV_max) was based on body weight and was calculated using a volume of interest that was drawn to encompass the whole 3D extent of the lesion. SUV_max was then calculated using the following equation: SUV_max = A/(ID/BW), where A is the maximum decay-corrected activity concentration in tissue/volume of interest (measured in millicuries per milliliter), ID is the injected dose of [18F]fluorodeoxyglucose (measured in millicuries), and BW is the patient's body weight (measured in grams). This gives SUV_max units in grams per milliliter. Changes in SUV (ΔSUV) after treatment were calculated with the following equation: ΔSUV = ([SUV_pre − SUV_post]/SUV_pre) · 100, where SUV_pre and SUV_post denote pre- and post-treatment SUV, respectively. At our institution, we follow the National Cancer Institute guidelines for image preparation, acquisition, and analysis.32-34 We used the maximum SUV because it is reproducible, unlike mean SUV, and reporting of maximum SUV is the standard at our institutions. We recognize that optimum SUV assessment is currently lacking. Baseline and postchemoradiation SUVs were compared.

The time interval between the baseline PET-CT scan and preoperative PET-CT scan was 12 ± 2 weeks.

Statistical Methods

Logistic regression analysis35 was used to determine the associations between pathologic complete response and percentage SUV decrease. Survival analyses were performed for OS and event-free survival (EFS). OS was defined as the time period from surgery to death or last follow-up, whichever occurred first. Patients who were alive at the time of the last follow-up were treated as censored. EFS was defined as the time period from surgery to death, recurrence, or last follow-up, whichever occurred first. Patients who were alive without recurrence at the time of the last follow-up were treated as censored. The Kaplan-Meier36 method was used to estimate the probability of OS and EFS. Log-rank test37 and Cox proportional regression analysis38 were performed to determine the association of initial SUV with OS and EFS. All statistical analyses were carried out using Splus 7.0 (Insightful Corporation, Seattle, Wash).39

RESULTS

Patient Characteristics

This was a retrospective analysis. Pertinent patient and treatment characteristics are listed in Table 1. All 151 patients had an adenocarcinoma of the esophagus or gastroesophageal junction. All patients received a fluoropyrimidine with concomitant radiation and another agent(s). A taxane was administered to 57% (n = 84) of patients, and a platinum compound was administered to 55% (n = 81) of patients. About 21% (n = 31) of patients had a complete pathologic response, and an additional 63% (n = 93) had some degree of response to chemoradiation (1% to ≤50% residual carcinoma in the resected specimen). Only 41 patients had induction chemotherapy before chemoradiation. The median follow-up exceeded 3.5 years, with maximum follow-up of 5.7 years. The interval between the end of chemoradiation and postchemoradiation PET/CT varied between 5 and 6.5 weeks.

Table 1. Patient and Treatment Characteristics

Descriptor	Frequency (%)
The median age was 61 years (range, 26-80 years) * Classified according to Siewert and Stein.30
Site of primary tumor*
Type I	106 (70)
Type II	40 (27)
Type III	5 (3)
Cytotoxins administered
Fluorouracil	151 (100)
Taxane	31 (21)
Platinum compound	58 (39)
Taxane + platinum compound	23 (16)
Taxane with others	30 (20)
Others	6 (4)
% residual cancer
0	31 (20)
1-50	93 (63)
>50	24 (16)

Percentage Decrease in SUV and Survival

The percentage decrease in SUV was dichotomized by 52%, based on a previous report.25 The log-rank test suggested that a >52% decrease in SUV after chemoradiation was highly associated with a longer overall survival time (P = .023; Table 2 and Fig. 1). Similarly, a larger fraction of patients who had a >52% decrease in SUV survived 3 years (72%) compared with those who had a ≤52% decrease in SUV (43%). The median survival of patients with ≤52% decrease in SUV was 2.49 years (95% confidence interval, 1.46 to not applicable), and the Kaplan-Meier method-estimated OS curve was >50% for those with a >52% SUV decrease. The EFS demonstrated a nonsignificant trend (P = .14) in favor of those who had a >52% decrease in SUV postchemoradiation (Fig. 2).

Figure 1. A Kaplan-Meier plot shows overall survival by percentage standardized uptake value decrease after chemoradiation (n = 151). E indicates number of events; N, denominator.

Figure 2. A Kaplan-Meier plot shows event-free survival by percentage standardized uptake value decrease after chemoradiation (n = 151). E indicates number of events; N, denominator.

Table 2. Overall Survival by Dichotomized Percentage Decrease in SUV Using Log-Rank Test

Parameter	Type	No.	Deaths, No.	Median Survival Time, y*	Rate of OS at 1 Year*	Rate of OS at 2 Years*	Rate OS at 3 Years*	P
SUV indicates standardized uptake value; NR, not reached. * Numbers in parentheses represent 95% confidence intervals.
% decrease in SUV	≤52	64	22	2.49 (1.46 to NR)	0.80 (0.69-0.92)	0.55 (0.41-0.74)	0.43 (0.28-0.64)	.02
	>52	87	17	NR (3.9 to NR)	0.83 (0.74-0.92)	0.79 (0.7-0.9)	0.72 (0.6-0.86)

A variety of parameters, including percentage decrease in SUV as a dichotomized and as a continuous variable, were evaluated by fitting the univariate Cox proportional hazards model (Table 3). The analysis for percentage decrease in SUV as a continuous variable indicated that the higher the decrease in SUV, the lower the risk of death (hazard ratio [HR], 0.99, P = .01). The analysis involving the dichotomized variable also confirmed this finding; patients who had a >52% decrease in SUV had a lower risk of death compared with those who had a ≤52% decrease in SUV (HR, 0.48; P = .026).

Table 3. Fitting Univariate Cox Proportional Hazards Models for Overall Survival, Using Continuous and Dichotomized Variables as Covariates

Variables	HR (95% CI)	P
HR indicates hazard ratio; CI, confidence interval; SUV, standardized uptake value. * Percentage SUV decrease used as continuous variable. † Percentage SUV decrease dichotomized by a cutpoint.
Initial SUV	0.96 (0.92-1.01)	.16
SUV postchemoradiation	1.05 (0.95-1.16)	.35
% SUV decrease postchemoradiation*	0.99 (0.988-0.998)	.01
% SUV decrease postchemoradiation†
≤52 (n=64)	1.00
>52 (n=87)	0.48 (0.26-0.92)	.026

Percentage Decrease in SUV and Pathologic Response

In the univariate analysis, the degree of residual cancer in the surgical specimen was highly correlated with survival (P = .003). Similarly, EFS was associated with the degree of pathologic response (P = .012). The percentage SUV decrease correlated marginally with pathologic complete response (univariate odds ratio [OR], 1.01; P = .06 and multivariate OR, 1.03; P = .07). Patients with chemoradiation-resistant cancer with >50% residual carcinoma in the resected specimen had a higher risk of death (HR, 2.12; P = .099, Table 4).

Table 4. Multivariate Cox Proportional Hazards Model for Overall Survival

Parameter	HR (95% CI)	P
HR indicates hazard ratio; CI, confidence interval; SUV, standardized uptake value. *Percentage SUV decrease used as continuous variable.
% residual cancer cells in the surgical specimen
0%-50%	1.00
>50%	2.12 (0.87-5.17)	.099
% decrease in SUV after chemoradiation*	0.99 (0.987-0.998)	.01

However, the multivariate analysis indicated that percentage decrease in SUV after chemoradiation was the only significant prognosticator (P = .01) of OS, and pathologic response was marginally associated with OS (P = .09).

DISCUSSION

In patients with localized carcinoma of the esophagus who undergo chemoradiation followed by surgery, subsequent development of metastases remains a major threat and results in frequent mortality. The molecular mechanisms of metastatic progression remain poorly understood.40, 41 However, increasing technological sophistication leading to better understanding of cancer's molecular biology suggests that complex genetic and epigenetic alterations are responsible for diverse clinical biology of cancers.10-12 Much work needs to be done before patients' outcome can be assuredly determined by studying the molecular biology, but the need to provide optimum therapy for cancer patients is acute. Some form of investigation that would classify tumors having good or poor biology is desired over empirical methods of treating every patient with a similar strategy, because the therapy has considerable morbidity and mortality. Only PET has been able to provide more discrimination that other clinical staging procedures.20, 21, 24, 25, 27, 42 The future of imaging with more specific tracers appears promising.43

By assessing the uptake of a radioactive glucose analog, PET/CT provides functional information about the cancer, most likely reflecting the biologic behavior of a particular cancer in terms of proliferation, metastatic potential, sensitivity to therapy, and more. Considerably more research is necessary to understand various aspects. For esophageal cancer, percentage changes in SUV have been studied after chemotherapy and chemoradiotherapy. Although the preliminary results suggest that PET/CT can potentially be a prognosticator for OS, the data on meaningful prediction of response are lacking. By this we mean that a large number of patients have not been studied and prospective studies are not reported.

We now know that patients with localized carcinoma of the esophagus can be separated into 3 distinct categories: 1) pathologic complete response, 2) some response (1% to ≤50% residual cancer cells in the resected specimen), and 3) no response (>50% residual cancer cells in the specimen).18, 31 These categories are not arbitrary, and not only can they be reproduced31 but they also impact OS18 and dictate metastatic progression.16 The most interesting aspect is that percentage SUV changes in PET/CT seem to narrow these 3 categories into 2 categories prior to surgery: 1) some response (this includes complete pathologic response and 0% to ≤50% residual cancer cells) and 2) no response (>50% residual cancer cells).23, 28 This should be considered an advance; however, PET/CT changes have not been able to discriminate between those with pathologic complete response and those with some residual cancer. Unfortunately, PET/CT changes do not correlate with pathologic complete response.23, 24, 28, 44-47

Some limitations of our results must be considered. This is a retrospective analysis; the dichotomized cutoff value of >52% was chosen from a study of squamous cell carcinoma, but it certainly performed well for adenocarcinoma. Other cutoff values could also be important, and correlations between the SUV changes and the degree of pathologic response remain nonsignificant, probably because of the relatively small number of patients in our study (although it is the largest yet reported).

Our data, which represent 1 of the largest series of gastroesophageal adenocarcinoma reported thus far, suggest that using percentage SUV changes postchemoradiation as a continuous variable provides a more accurate reflection of the clinical biology of esophageal cancer than the literature-suggested method of using dichotomized results. We also detected a nonsignificant trend for the identification of pathologic complete response in the univariate and multivariate analyses.

In conclusion, the percentage SUV change in PET/CT after chemoradiation, when used as a continuous variable, is an independent prognosticator of OS and can classify gastroesophageal adenocarcinoma patients into 2 groups: 1) pathologic complete response and some response; and 2) no response; however, to detect significant associations a larger number of patients must be studied. PET/CT is likely to complement other tools to individualize therapy for patients with localized gastroesophageal adenocarcinoma.

Acknowledgements

We thank Drs. Alexandria Phan, Ritsuko Komaki, Alexander Dekovich, Jeremy Erasmus, David Rice, William A. Ross, and Ara Vaporciyan for invaluable contributions.

Conflict of Interest Disclosures

Supported in part by the Dallas, Cantu, Smith, and Park Families and by the Rivercreek Foundation.