Decreased Posttreatment SUV on PET Scan Is Associated With Improved Local Control in Medically Inoperable Esophageal Cancer

ABSTRACT

Background:

The relationship between local, regional, or distant disease control (LC, RC, DC) and maximal posttreatment standardized uptake value (SUV_max) in patients with esophageal cancer has not been elucidated. This study was initiated to explore whether a decrease in SUV on positron emission tomography-computed tomography (PET-CT) scan is associated with LC, RC, or DC in patients with esophageal carcinoma treated with definitive chemoradiotherapy.

Methods:

Medical records of 40 patients with inoperable esophageal cancer treated with definitive intent and who underwent pre- and posttreatment PET-CT scans were reviewed. The histology, nodal status, tumor location, and radiotherapy (RT) dose were investigated as variables to determine a relationship between SUV_max and LC, RC, and DC as well as disease-free survival (DFS).

Results:

Decreased posttreatment SUV_max on PET scan (P = .02) and increased RT dose (P = .009) were the only significant predictors of improved LC on univariate analysis. Mean RT doses in patients with no evidence of disease or with local, regional, or distant recurrences were 5,244, 4,580, 5,094, and 4,968, respectively. Decreased posttreatment SUV (P = .03) and increased RT dose (P = .008) were also associated with an improvement in DFS. Furthermore, decreased posttreatment SUV_max correlated with an improvement in LC (hazard ratio [HR] = 1.3, 95% confidence interval [CI] = 1.03–1.6, P = .03) as well as DFS (HR = 1.3, 95% CI = 1.03–1.6, P = .03). These findings were maintained on multivariate analysis.

Conclusions:

Posttreatment decrease in SUV is associated with LC and DFS in esophageal cancer patients receiving definitive chemoradiotherapy. RT dose was also associated with both LC and DFS. The prognostic significance of these findings warrants prospective confirmation.

An estimated 16,470 new esophageal cancers will be diagnosed in the United States in 2008, accounting for approximately 14,280 deaths.¹ The extremely high rate of mortality can be attributed to the disease presenting in a locally advanced or metastatic state at diagnosis. Optimal management of locally advanced esophageal cancer remains an area of debate. Recent reports have confirmed the therapeutic superiority of trimodality treatment over other options, as demonstrated by operative pathology and local control.^2–4 In medically inoperable cases, or if the patient refuses surgical intervention, local failure rates as high as 50% have been reported with concurrent chemoradiotherapy alone.⁵ It has been suggested that radiotherapy dose escalation (as part of combined-modality chemoradiotherapeutic treatment) may yield superior disease control rates,⁶ though a subsequent randomized trial was inconclusive, due to early study termination.^7,8

Assessment of tumor response to therapy using positron emission tomography-computed tomography (PET-CT) has been proposed and supported in various malignancies.^9–12 In the case of the esophageal cancer specifically, Weber et al¹³ demonstrated that an absence of metabolic response on PET scans 14 days after induction chemotherapy in 40 patients with adenocarcinoma of the esophagus portended a shorter time to progression/recurrence and shorter overall survival. Lordick and coworkers¹⁴ suggested using PET response to tailor and individualize multimodality treatment.

We have previously demonstrated an association with disease-free survival, a trend toward an association with a complete pathologic response, and posttreatment PET maximum standard uptake value (SUV_max).¹⁵ However, the prognostic role of PET-CT imaging in the nonsurgical management of esophageal cancer remains unclear.

This retrospective analysis addressed two specific hypotheses. The first was to determine whether a decrease in SUV_max on a PET-CT scan was associated with local, regional, or distant control in esophageal carcinoma patients treated with definitive chemoradiotherapy. We hypothesized that a greater response on post–combined-modality therapy PET would be associated with an improvement in one or more of the desired treatment outcomes. A secondary goal was to assess the effect of the RT dose (as part of combined-modality treatment) on the outcomes outlined above. We hypothesized that an increased RT dose would result in an improvement in treatment-related outcomes.

The clinical relevance of these findings would be twofold. Assessment of response in patients unable to undergo surgical resection would tailor any possible further chemotherapy alone or chemoradiotherapeutic interventions to the patient's needs, with the goal of improving local and potentially distant control. The second would be to identify patients who might benefit from higher RT doses.

MATERIALS AND METHODS

A retrospective study was performed on patients diagnosed with nonmetastatic carcinoma of the esophagus between January 2002 and June 2007. Patients who underwent surgery as part of the definition plan were intentionally excluded. Standard workup included esophageal ultrasonography and CT scanning, in addition to PET- CT scanning. Endoscopic ultrasound was not performed if it was not felt to contribute to the overall management of the patient. The PET-CT scans were usually obtained within a few days of the CT simulation, prior to initiation of any cytotoxic therapy, and were repeated 4–6 weeks posttreatment completion. Our PET-CT procedures have been previously published.¹⁶ Maximal SUVs were determined and recorded from PET –CT scan images by one nuclear medicine physician (JQY) who was blinded to the clinical outcome.

Our RT techniques have also been previously outlined.^15,16 Briefly, all patients underwent CT simulation in an immobilization device. Subsequently, PET scan images were fused with the CT simulation images to assist in the determination of the gross tumor volume and planning tumor volume. Patients were initially treated with anteroposterior and posteroanterior fields with 6- or 10-MV photon beams. An anteroposterior and two posterior oblique fields were incorporated into the treatment to limit the spinal cord to no more than 4,500 cGy. Customized blocks were used to protect normal tissue. The usual field borders were 5 cm superiorly, 3 cm inferiorly (unless the tumor was in the midesophagus, where the border was 5 cm inferiorly), and 2.5–3 cm radially to the gross tumor volume as outlined on the CT and PET scans. The chemotherapy regimen was left to the discretion of the treating medical oncologist.

Survival analysis was performed from the start of treatment. Patients dying of non–cancer-related causes after completion of therapy without evidence of cancer were censored at time of death. Local failure was defined as any failure within the radiation portal, either on posttreatment imaging or on rebiopsy. Regional failures were defined as failures included in the draining regional lymph node area, and distant failures were those occurring at remote sites. Local, regional, or distant control as well as disease-free survival (DFS) was recorded and correlated with histological type, nodal status, tumor location, and radiotherapy (RT) dose.

Univariate analyses to determine factors affecting local regional or distant control were performed using Cox proportional hazard modeling (for continuous variables) and Kaplan-Meier estimation method (for discrete variables).^17,18 Continuous variables analyzed included pre- and posttreatment SUV_max, percentage SUV_max change, RT dose, and age. Discrete variables included heart disease, diabetes, insulin levels, smoking history, histology, tumor location, T-stage, nodal status on endoscopic ultrasound, and chemotherapy administration.

Multivariate analyses were performed using the Cox proportional hazard model.¹⁹ A stepwise approach was used, and only those factors found to be significant on univariate analyses were included in the multivariate analysis model. These included posttreatment SUV decrease, RT dose, and chemotherapy administration. Posttreatment SUV_max was analyzed as both a continuous and a dichotomous event. All analyses were carried out using the SAS software package (SAS Software, Cary, NC).

All research related to this project was approved by the Institutional Review Board at Fox Chase Cancer Center. All patient identifiers were removed before analysis of the data.

RESULTS

Pretreatment patient characteristics are outlined in Table 1. Forty patients (68% men, 32% women) with biopsy-proven esophageal cancer were treated with a definitive nonsurgical approach between January 2002 and June 2007. Pre- and posttreatment PET-CT scanning data were available for this entire selected group of patients. Endoscopic ultrasonography was performed on 28 (76%) patients for determination of T-stage, with 3 (11%) having T2, 23 (82%) T3, and 2 (7%) T4 disease. Twenty-one patients had ultrasound- or CT-detected nodal disease. Adenocarcinoma, pathologically confirmed in 26 (65%) patients, was the most common histology, and 14 patients (35%) had squamous cell cancer.

Table 1.

Pretreatment patient characteristics

Sex
Men	68%
Women	32%
Location
Cervical	5%
Middle	35%
Lower	33%
GE junction	27%
T-stage
T2	11%
T3	82%
T4	7%
N-stage
N0	43%
N1	57%
Comorbidities
Smoking	74%
Cardiovascular disease	54%
Diabetes	24%

Tumor location was fairly evenly distributed, with two patients (5%) having cervical esophageal, 14 patients (35%) having middle esophageal, 13 (33%) having lower esophageal, and 11 (27%) having GE junction tumors. Five (12%) of the patients in our analysis did not undergo concurrent chemotherapy, at the discretion of the treating medical oncologist, due to a significant history of cardiovascular disease or patient refusal to undergo chemotherapy being the primary causes.

Treatment and follow-up details are outlined in Table 2. Median age at initiation of treatment was 75 years (range 43–97). Median RT dose was 5,040 cGy (range 720–6,208 cGy). Median pretreatment tumor SUV_max on PET CT scanning was 9.5 (0–33), and the corresponding median posttreatment value was 3 (0–16.3). Median percent change in SUV_max posttreatment was −65% (range, −91%–254%). Mean follow-up was 10.4 months (range, 0.47–52.83).

Table 2.

Treatment and followup characteristics

Variable	Mean	Median	Minimum	Maximum
Age (years)	72.15	75	43	97
RT dose (Gy)	5,032.5	5,040	720	6,208
Initial SUVmax	10.23	9.5	0	33
Posttreatment SUVmax	3.91	3	0	16.3
% SUVmax decrease	−35.79	−64.94	−91.21	254.35
Followup (months)	10.42	5.78	0.47	52.83

Abbreviations: RT = radiotherapy; SUV_max = maximum standard uptake value.

Twenty patients (50%) had no evidence of disease (NED) at the time of study analysis. Recorded failures were local in 9 (23%), regional in 6 (15%), and distant in 5 (13%). Mean RT doses in patients with local, regional, or distant recurrences were 4,580, 5,094, and 4,968, respectively. Mean RT dose in patients without evidence of disease was 5,244 cGy.

A decrease in posttreatment SUV on PET scan (P = .02) as well as increased RT dose (P = .009) were the only significant predictors of improved local control on univariate analysis. There were no significant variables associated with regional or distant control. Furthermore, posttreatment decreased SUV_max on PET (P = .03) and increased RT dose (P = .008) were associated with an improvement in DFS. A decrease in posttreatment SUV_max also correlated with an improvement in local control (hazard ratio [HR] = 1.3, 95% confidence interval [CI] = 1.03–1.6, P = .03) as well as DFS (HR = 1.3, 95% CI = 1.03–1.6, P = .03) on multivariate analysis. No other significant factors were identified for local, regional, or distant control or DFS.

DISCUSSION

The nonsurgical chemoradiotherapeutic treatment of medically inoperable esophageal cancer presents a challenge in determining appropriate response and consideration for further therapy. RTOG 85–01 compared radiotherapy alone (to 64 Gy) versus a combined approach of 5-fluorouracil (5-FU) and cisplatin with external beam radiotherapy (to 50 Gy). The combined-modality arm showed superior survival (26% 5-year survival), establishing this regimen as the definitive approach of choice in the nonsurgical management of esophageal cancer.^20–22

The first two chemotherapy cycles were administered concurrently with radiotherapy (during the first and fifth weeks). Persistence of local disease remained the most common cause of treatment failure. A retrospective review by Liao and coworkers of 132 patients with stage II or III esophageal cancer treated with chemoradiotherapy alone or followed by surgery demonstrated significant improvement in locoregional control, DFS, and overall survival rates in patients undergoing postchemoradiotherapy esophagectomy.²³

Several randomized trials have also demonstrated the benefits of trimodality treatment.^2–4,24 This has stimulated interest in determining which inoperable patients may benefit from further nonsurgical intervention following standard chemoradiotherapy. Given the high toxicity rates noted on the RTOG 85–01 trial, a functional imaging technique to determine responsiveness to induction treatment would be an ideal mode for determining success of treatment in the inoperable population.

In this regard PET scanning has been shown to be predictive of response following chemotherapy,^25,26 chemoradiotherapy,^27,28 and radiotherapy alone²⁹ for many different tumors, most notably lung cancer, head and neck cancer,³⁰ and gastric cancer.^31,32 In the case of esophageal carcinoma, most data regarding PET-CT scanning relate to initial tumor staging and metastatic workup.^33–35 Instances where PET data have been used to evaluate response to chemoradiotherapy have been in the setting of neoadjuvant therapy in preparation for surgical resection.^{13,34,36–39} Greater complete or subtotal tumor regression was reported in 44% of patients with a metabolic response, compared with only 5% without a metabolic response. Patients demonstrating an absence of metabolic response on posttreatment PET had a shorter time to disease progression and decreased overall survival. Furthermore, we previously presented data that suggested that posttreatment SUV_max after preoperative chemoradiotherapy in a surgical population correlated with a pathologic response noted in surgical specimen and could be predictive of disease-free survival. We also noted a correlation with pretreatment SUV_max and depth of tumor invasion.¹⁵ Our current study addressed the validity of these findings in the medically inoperable population and attempted to assess the role of posttreatment PET-CT SUV_max decrease as a surrogate for clinical response.

The studies noted above yielded interesting and often conflicting outcomes. Although Flamen, Swisher, and Wieder^34,38,39reported that the posttreatment scan corresponded with disease control and survival, a recent surgical series by McLoughlin and coworkers⁴⁰ looking at pathologic correlates of PET response concluded that a complete PET response to neoadjuvant chemoradiotherapy does not necessarily translate into a histologically superior outcome. Similar results were reported by Smithers and coworkers,⁴¹ who concluded that a fluorodeoxyglucose (FDG)-PET scan 3–6 weeks after completion of neoadjuvant chemoradiotherapy should not be used to assess response.

It is notable that both of these studies were in a neoadjuvant setting in a population otherwise earmarked for trimodality treatment. In the absence of the ability to obtain a pathologic confirmation of response, there has been emerging support of the use of posttreatment PET scanning as an indicator of treatment response. Our findings support the use of posttreatment PET-CT scanning at 4–6 weeks after chemoradiotherapy as an indicator of response. Despite a relatively small sample size, a posttreatment decrease in SUV_max was significantly associated with improved local control and disease-free survival on both univariate and multivariate analysis.

The optimal timing of posttreatment PET scan has also been questioned by Smithers and colleagues, especially in light of potential alteration of the SUV_max reading due to the inflammatory response associated with chemoradiotherapy. This issue has been previously addressed to a limited extent.^15,42 In this regard it is important to look at the SUV_max not globally but specifically, in the solid component seen on the CT. Wieder et al performed PET-CT scans in mid treatment and showed that inflammation of the normal esophagus was not a significant problem in measuring the tumor response: Diffuse esophageal FDG uptake suggesting esophagitis was observed in only a small number of patients (15% of patients after 14 days of chemoradiotherapy). The researchers concluded that the significant correlation between response and midtreatment FDG-PET metabolic activity might be useful in stratifying patients early in treatment and appropriately modifying the therapeutic approach.

In our study all posttreatment PET scans were invariably performed within a 2-week window between 4 and 6 weeks after chemoradiotherapeutic treatment. With inflammatory changes so prevalent, it is imperative for the interpreting and treating physicians to review the images to define the solid and inflammatory components and their corresponding SUV_max in relationship to the distribution of the treatment to minimize the possibility of false-positive readings. In our study interpretation of the PET scans by a single nuclear medicine physician (JQY) allowed for consistency in result interpretation as well as the individualized specific analysis of solid CT components differentiating them from an inflammatory background that may otherwise cloud SUV_max interpretation. This avoided physician bias in determining SUV_max changes.

An unexpected finding in this study was the association between higher RT dose and improved outcome. Previous studies attempting dose escalation in esophageal cancer have not been supportive of this strategy. Indeed, the RTOG 94–05 study, which attempted to compare chemotherapy plus concomitant RT with either 64.8 Gy versus 50.4 Gy, was terminated early due to increased toxicity in the higher dose arm. Interestingly, these toxicities were largely encountered at median doses lower than 50 Gy and, therefore, did not support the notion that they were associated with dose-escalated regimens. The outcomes were similar in both study arms (locoregional disease persistence 52% vs. 56%) at time of publication, though these results may have been affected by the premature termination of the study.

A previous phase II study at Fox Chase Cancer center by Coia and coworkers demonstrated improved local control (but no improvement in overall survival) with RT to 60 Gy in 2-Gy fractions in conjunction with chemotherapy.⁶ A subsequent small study by Sauter and coworkers, also at Fox Chase Cancer Center, using preoperative high-dose chemoradiotherapy to 60 Gy, had shown promising improvements in complete pathologic response local control rates following resection.⁴³

In our current study, increased RT dose was strongly associated with an improvement in local control as well as disease-free survival on univariate analysis, though this was reduced to a trend only on multivariate analysis. The 21 patients who were classified as NED at time of manuscript preparation received a median RT dose of 5,244 cGy compared with 4,580 cGy in patients with local failure, 5,094 cGy in patients with regional failure, and 4,968 cGy in those with distant failure (P = .009). However, it is critical to note that this dose-response relationship is within the higher limits of the standard treatment range and differs from the aforementioned studies that attempted to escalate doses beyond this range. Figure 1 represents a Kaplan-Meier representation of local recurrence based on a cut-point of greater than or less than 50 Gy RT dose. Once again, the caveat of low numbers in each group persists. However, we believe that the data are hypothesis generating.

Figure 1.

Kaplan-Meier representation of local recurrence by RT dose ≤50 Gy or >50 Gy.

In summary, we conclude that based on a single posttreatment FDG-PET at 4–6 weeks, when compared with a pretreatment PET, a drop in the postchemoradiotherapy SUV_max is associated with improvements in local control and DFS. The absence of posttreatment SUV_max decrease appeared to portend a worse prognosis in our treatment group, and this population of patients may benefit from further systemic treatment. Furthermore, these data support efforts to maintain target radiation doses in the nonsurgical definitive treatment of esophagus cancer. These conclusions are limited by the retrospective nature of this study and the small sample size. However, they provide a basis for further prospective evaluation of PET scanning as a prognostic indicator in esophagus cancer treatment.