Abstract
Objectives
We compared the effectiveness of upfront esophagectomy versus induction chemoradiation followed by esophagectomy for overall survival in patients with clinical T2N0 esophageal cancer. We also assessed the impact of the diagnostic uncertainty of endoscopic ultrasound on the expected benefit of chemoradiation.
Methods
We created a decision analysis model representing two treatment strategies for cT2N0 esophageal cancer: 1) upfront esophagectomy which may be followed by adjuvant therapy for upstaged patients and 2) induction chemoradiation for all cT2N0 patients followed by esophagectomy. Parameter values within the model were obtained from published data, and median survival for pathologic subgroups was derived from the National Cancer Database. In sensitivity analyses, staging uncertainty of endoscopic ultrasound was introduced by varying the probability of pathological upstaging.
Results
The baseline model showed comparable median survival for both strategies: 48.3 months for upfront esophagectomy versus 45.9 months for induction chemoradiation and surgery. The sensitivity analysis demonstrated induction chemoradiation was beneficial with probability of upstaging greater than 48.1%, which is within the published range of 32–65% probability of pathologic upstaging after cT2N0 diagnosis. The presence of any of three key variables: size over 3 cm, high grade, or lymphovascular invasion was associated with greater than 48.1% risk of upstaging, thus conferring a survival advantage to induction chemoradiation.
Conclusions
The optimal treatment strategy for cT2N0 esophageal cancer depends upon the accuracy of endoscopic ultrasound staging. High-risk features that confer increased probability of upstaging can inform clinical decision-making to recommend induction chemoradiation for select cT2N0 patients.
INTRODUCTION
Esophageal cancer is a common malignancy in the United States, with approximately 17,000 new cases annually.1 Disease stage is based on the American Joint Committee on Cancer (AJCC) criteria2 and determines the treatment paradigm. Treatment for early stage, localized disease (T1N0) is primary surgery with no proven benefit from induction therapy, and possibly a detriment in outcomes.3,4 Patients with locally advanced disease (T2–4, N1–3) have improved outcomes with induction chemoradiotherapy compared to surgery alone.5 There is conflicting data from retrospective studies on the benefit of neoadjuvant chemoradiation for clinical T2N0 (cT2N0) disease.4,6–9 This observed variability in results may be due to inaccuracies in clinical staging.
Endoscopic ultrasound (EUS) is the key diagnostic tool used for determining clinical stage in locoregional disease by assessing the depth of tumor invasion (T) and presence of regional lymphatic spread (N). Clinical EUS findings have been compared to pathologic staging in surgical resection specimens and accuracy varies drastically by stage.10 In patients with early and locally advanced disease, EUS has a sensitivity of 75% and 73.7%, respectively, for properly identifying the stage of disease.10 In patients with cT2N0 disease, EUS performs poorly with a sensitivity of only 9.1%.10 Several series have found that cT2N0 patients truly had pathologic T2N0 (pT2N0) disease only 6–29% of the time, with probabilities of a true higher stage ranging from 32–65%.4,8–16 These pathologically upstaged patients were actually understaged clinically with EUS. Since clinical staging is crucial in selecting appropriate treatments to maximize survival, diagnostic uncertainty in cT2N0 esophageal cancer has led to controversy about the optimal treatment strategy. Consequently, clinical practice with regards to providing induction therapy versus upfront surgery is equivocal.8,16
The gold standard for determining the optimal treatment for cT2N0 patients might be to conduct a randomized controlled trial. Conducting such a large scale trial in this limited subset of esophageal cancer patients would be costly, accrue patients slowly, and require long-term follow up, making a clinical trial a very difficult way to answer this question. Therefore, to address this problem, we created a decision analysis model to evaluate the role of neoadjuvant therapy versus upfront resection in cT2N0 esophageal cancer patients and to identify the threshold for benefit of induction chemoradiation. We explored the effect of the diagnostic uncertainty of EUS and operative risk on the model, and evaluated tumor characteristics that affect a patient’s probability of upstaging.
METHODS
Decision Analysis Model
To determine the optimal strategy for treatment of patients with cT2N0 esophageal cancer, we developed a decision analysis model using TreeAge Pro software (2016, Version R2.1, Williamstown, MA). The model compared survival based on initial treatment with surgery or chemoradiation. Decision analysis uses mathematical modeling to compare different treatment modalities by modeling plausible outcomes over a range of clinical scenarios, estimating the effectiveness of each treatment strategy. Sensitivity analyses can be performed to incorporate uncertainty and assess the driving factors in the model. The probabilities and median survival used to populate the decision tree were obtained from published literature and a large validated database, the National Cancer Database (NCDB).
Treatment Strategies and Structure of the Model
We modeled two groups of cT2N0 patients eligible for surgery: 1) upfront esophagectomy which may be followed by adjuvant therapy for pathologically upstaged patients and 2) induction therapy followed by esophagectomy. The model evaluated which treatment strategy would yield a better overall survival. These two general treatment paradigms are reflected in our decision tree (Figure 1). Patients in the esophagectomy arm were dichotomized to perioperative death (defined as death within 30 days of surgery) or survival. Patients surviving esophagectomy may receive adjuvant systemic therapy if they have locally-advanced disease (T3 or N+). In contrast, patients in the induction chemoradiation arm may experience death related to chemoradiation, become a non-operative candidate due to treatment-related morbidity or progression of disease, or they may remain an operative candidate and proceed to esophagectomy.
Figure 1.
Decision tree comparing upfront esophagectomy to induction chemoradiation followed by esophagectomy.
Model Parameters and Data Sources
The baseline values used in the model are described in the following paragraphs and are summarized in Table 1.
Table 1.
Probabilities used in the baseline decision-analysis model.
| Variable Description | Data Source | Value |
|---|---|---|
| Upfront Esophagectomy | ||
| Perioperative death | Hofstetter et al (2014)13 | 0.03 |
| Probability of patient upstaging | Speicher et al (2014)8 | 0.416 |
| Probability of pT2N0 | Speicher et al (2014)8 | 0.267 |
| Probability of patient downstaging | Speicher et al (2014)8 | 0.317 |
| Probability of receiving adjuvant therapy if upstaged | Speicher et al (2014)8 | 0.50 |
| Survival with adjuvant therapy in upstaged patients | Samson et al (2016)16 | 34.6 months |
| Survival with surgery only in upstaged patients | Samson et al (2016)16 | 20.8 months |
| Survival with surgery only in pT2N0 patients | Derived from NCDB | 44.4 months |
| Survival with surgery only in downstaged patients | Derived from NCDB | 83.6 months |
| Induction Chemoradiation followed by Esophagectomy | ||
| Probability that patient becomes a non-operative candidate after induction therapy | van Hagen et al (2012)5 | 0.06 |
| Chemoradiation-related death | van Hagen et al (2012)5 | 0.006 |
| Probability of post-induction higher pathological stage | Speicher et al (2014)8 | 0.341 |
| Probability of post-induction ypT2N0 | Speicher et al (2014)8 | 0.243 |
| Probability of post-induction lower pathological stage | Speicher et al (2014)8 | 0.416 |
| Post-induction perioperative death | Hofstetter et al (2014)13 | 0.03 |
| Survival if patient has a perioperative death after induction | van Hagen et al (2012)5 | 2.3 months |
| Survival if patient becomes a non-operative candidate | Abrams et al (2009)18 | 16.0 months |
| Survival with trimodality therapy in upstaged patients | Derived from NCDB | 34.6 months |
| Survival with trimodality therapy in ypT2N0 patients | Derived from NCDB | 43.6 months |
| Survival with trimodality therapy in downstaged patients | Derived from NCDB | 66.5 months |
Perioperative Mortality
Perioperative outcomes were assessed using 30-day mortality estimated at 3%13 in both the upfront esophagectomy and induction chemoradiation arms (Figure 1). In patients who died within the perioperative period after upfront surgery, survival was counted as 0 months. However, patients receiving induction chemoradiation have additional survival time during neoadjuvant chemoradiation prior to perioperative death at time of esophagectomy. Patients dying in the perioperative period after induction therapy were assigned a median survival of 2.4 months to account for chemoradiation therapy and the standard delay following induction therapy before surgery based on the findings in the CROSS trial.17
Morbidity and Mortality of Induction Therapy
Mortality during induction chemoradiation in our baseline model was 0.6% based upon the CROSS trial.17 Patients who had a death related to induction therapy were assigned a survival of 0 months. Patients receiving induction therapy may not be operative candidates after completion of therapy. Based upon CROSS trial data, we used an estimate of 6% in our baseline model to represent the proportion of patients who underwent chemoradiation and were no longer operative candidates because of progression of disease, patient preference, toxicity of chemotherapy, or decline in health making them medically unfit for surgery.17 Median survival for patients who were deemed inoperable after induction therapy was 16 months.18
Accuracy of Clinical Staging
In order to reflect the uncertainty in EUS and the impact on survival in the esophagectomy arm, we included the probability of upstaging, downstaging, and accurate staging using pathological data in patients who had undergone upfront esophagectomy. A literature search revealed ten studies that examined the diagnostic accuracy of endoscopic ultrasound in cT2N0 cancers. A summary of these studies, the number of patients included, and the proportions of patients with same, down-, or upstaged disease is included in Table 2. For our baseline analysis, the largest study by Speicher et al8 was used, and the probabilities of up-, same-, and downstaging were set at 0.416, 0.267, and 0.317 respectively. As shown in the table, the range of upstaging probability ranged from 32–65%, and this variability was used as the basis for our one-way sensitivity analysis of diagnostic uncertainty (Figure 1). As we varied the probability of upstaging in our model from 0.3 to 0.7 to encompass the entire published range, we held the ratio of the remaining proportion of same- and downstaged patients at 1:1. This reflected the ratios seen in the largest clinical studies8,12 and prevented an unrealistic elimination of a pathologic subgroup at the high end of the sensitivity analysis.
Table 2.
Published diagnostic accuracy of endoscopic ultrasound in patients staged as clinical T2N0 cancer undergoing upfront surgery. In patients undergoing upfront surgery, pathological upstaging is equivalent to EUS understaging and means that the patient was found to have T3/4 or N+ disease following surgical resection. Pathologic downstaging is equivalent to EUS overstaging and means the patient had T1N0 disease.
| Study | N | Probability of pT2N0 |
Probability of Pathologic Downstaging |
Probability of Pathologic Upstaging |
|---|---|---|---|---|
| Rice et al (2007)4 | 53 | 13.0% | 55.0% | 32.0% |
| Crabtree et al (2011)10 | 18 | 6.0% | 50.0% | 44.0% |
| Stiles et al (2011)11 | 40 | 12.5% | 22.5% | 65.0% |
| Zhang et al (2012)9 | 14 | 29.0% | 21.0% | 50.0% |
| Crabtree et al (2013)12 | 752 | 27.4% | 25.9% | 46.7% |
| Hofstetter et al (2014)13 | 499 | 14.0% | 44.0% | 42.0% |
| Hardacker et al (2014)14 | 68 | 8.5% | 42.8% | 48.5% |
| Speicher et al (2014)8 | 1599 | 26.7% | 31.7% | 41.6% |
| Tekola et al (2014)15 | 38 | 21.0% | 47.0% | 32.0% |
| Samson et al (2016)16 | 932 | 54.3% | 45.7% | |
| Range | 6–29% | 21–55% | 32–65% | |
Treatment Effect of Induction Therapy
To evaluate the treatment effect of induction therapy, pathological stage after trimodality treatment with induction chemotherapy and radiation followed by esophagectomy was included in our model. For our baseline analysis, the largest study by Speicher et al8 was used, and the probabilities of upstaging, same-stage, or downstaging for the induction therapy arm were set at 0.341, 0.243, and 0.416 respectively. These probabilities reflect the both the diagnostic uncertainty of EUS as well as the treatment benefit of induction therapy, as fewer patients are likely to be upstaged compared to upfront surgery. When conducting our sensitivity analysis of diagnostic uncertainty, we held the ratio of patients upstaged after upfront surgery to patients upstaged after induction therapy and surgery constant at 0.82, based on the ratio observed in the study by Speicher et al.8 Additionally, we held the ratio of the remaining patients with the same stage versus downstaged stable at 1:2. This incorporated the expected treatment effect of chemoradiation in a realistic way, preventing distortion of the probabilities of each outcome as the chance of upstaging was varied.
Adjuvant Therapy Use
Patients with cT2N0 who undergo upfront esophagectomy and are upstaged are candidates for adjuvant chemoradiation. Practice patterns for the use of adjuvant therapy are inconsistent. We used data from multiple NCDB studies which demonstrated that, in current practice, approximately 50% of upstaged cT2N0 patients receive adjuvant therapy.8,16
Survival Outcomes for Patients in Each Arm
For upstaged patients undergoing upfront esophagectomy, median survival data was included from a study by Samson et al16, which explicitly examined upstaged cT2N0 patients stratified based on receipt of adjuvant therapy. Those receiving adjuvant therapy had a median survival of 34.6 months vs 20.8 months for surgery alone, and these values were incorporated into our model. For the remaining arms, no published data was available that identified median survival for individual groups of cT2N0 patients treated with upfront surgery or induction chemoradiation separated by their ultimate pathologic stage. Therefore, we used data from the NCDB to generate Kaplan Meier Survival curves using SAS software (Version 9.4, SAS System for Windows, Copyright © 2012, SAS Institute Inc., Cary, NC, USA). The survival curves can be seen in Figure 3 (supplementary image). Median survival estimates resulting from these curves were: 44.4 months for pT2N0 disease after upfront surgery, 83.6 months for downstaged patients after upfront surgery, 34.6 months for patients pathologically upstaged despite receiving induction therapy, 43.6 months for ypT2N0 disease, and 66.5 months for downstaged patients after induction chemoradiation.
Supplementary: Figure 3.
Kaplan Meier Survival Curves. We primarily derived median survival by pathologic subgroup for (a) patients receiving induction chemoradiation and (b) patients undergoing upfront esophagectomy after clinical T2N0 diagnosis through use of Kaplan Meier analysis for input into our decision tree. Relevant median survivals that were included in our model are detailed in Table 1.
Decision Analysis and Sensitivity Analyses
The decision analysis model was run using TreeAge Pro to compute expected survival time for each treatment paradigm to determine which yields the longest expected length of survival, incorporating the evidence-based input probabilities for entering each branch point with median survival as the outcome of interest for the terminal node. All of the baseline characteristics described above are listed in Table 1.
Within decision analysis, sensitivity analyses assess the stability of the model over a range of probabilities, exploring how the assumptions made with the various baseline inputs affect the overall outcome of the model. One-way sensitivity analysis was conducted on the probability of upstaging in the esophagectomy arm as described above to incorporate the diagnostic uncertainty of endoscopic ultrasound. Additionally, we performed a one-way sensitivity analysis examining operative risk by varying the probability of perioperative mortality between 3% and 10%, a clinically plausible range.13,19,20 We also evaluated the effect of differential perioperative mortality between the two arms. Despite experienced centers reporting comparable mortality rates between the two strategies13, a meta-analysis estimated patients treated with chemoradiation may have twice the perioperative mortality as patients receiving surgery alone21, so we varied the mortality rate up to 6% in the induction therapy arm.
Identifying Features that Predict Upstaging
After identifying the threshold of EUS understaging at which a benefit would be expected from induction chemoradiation, we sought to investigate clinical or pathologic features of patients or their tumors that would place them at risk of pathologic upstaging higher than the identified threshold, suggesting these individuals might benefit from induction chemoradiation. To do this, we analyzed cT2N0 patients from the NCDB who underwent upfront surgery from 1998–2012. Descriptive statistics were performed comparing tumor characteristics (grade, histology, lymphovascular invasion, and tumor size) and patient characteristics (age, sex, race, urban/rural, cancer center type, insurance status, education status by zip code, income quartile by zip code, and Charlson Deyo score) by pathological staging group with Chi-Squared, Kruskal-Wallis, and ANOVA tests. Univariable and multivariable analyses were performed using binary logistic regression to assess the possible predictive value for each of these variable for upstaging versus same- or downstaging.
RESULTS
Baseline Analysis
Our model resulted in an expected survival of 48.3 months in the patients receiving upfront esophagectomy compared to 45.9 months in patients treated with induction chemoradiation, yielding an estimated survival benefit of approximately 2.5 months with the upfront surgery strategy. To restate in another way, routine application of induction therapy was likely to be harmful rather than beneficial with regards to survival.
Threshold Analysis for Endoscopic Ultrasound Diagnostic Uncertainty
The proportion of patients likely to be upstaged, and therefore benefit most from induction chemoradiation, across many published studies ranges from 32–65% as shown in Table 2. We varied this parameter in the one-way sensitivity analysis while holding the ratio of same and downstaged patients stable at 1:1. The results of this analysis (Figure 2) revealed that with a probability of upstaging of 48.1% or greater, induction chemoradiation would be expected to confer a survival benefit over upfront esophagectomy.
Figure 2.
One-way sensitivity analysis varying the probability of upstaging between 30–70%, with a threshold for benefit of induction chemoradiation seen at 48.1%.
Sensitivity Analysis for Perioperative Mortality
In another one-way sensitivity analysis, we varied the probability of perioperative mortality up to 10%, keeping all other inputs stable. The results showed that upfront esophagectomy is the favored strategy compared to neoadjuvant treatment with a survival benefit of 2–3 months throughout this range of surgical risk. When we evaluated differential risk in perioperative mortality between the two strategies by increasing the rate in the induction arm, we found that the threshold of pathological upstaging for benefit of induction chemoradiation increased to 50.8%, 53.4%, and 55.9% at 4%, 5%, and 6% mortality, respectively.
Identifying Features that Predict Upstaging
Tumor and patient characteristics were examined in 1520 NCDB patients who were cT2N0, underwent upfront surgery, and had available pathologic staging data. These characteristics were compared across patients who were pathologically downstaged, pT2N0, and upstaged and the results of selected descriptive comparisons can be seen in Table 3. Tumor grade (p<0.001), histology (p<0.001), lymphovascular invasion (p<0.001), tumor size (p<0.001), sex (p=0.008), cancer center type (p<0.007), and education status (p=0.013) were significantly different between the pathological staging groups. Age, race, urban/rural, insurance status, income quartile by zip code, and Charlson Deyo score were not significantly different between groups.
Table 3.
Characteristics of cT2N0 NCDB patients undergoing upfront surgery by true pathologic stage. Thick boxes highlight several tumor features that are associated with upstaging. Presence of these features place patients above the 48.1% probability of upstaging threshold where they may benefit from routine induction chemoradiation.
| Variable | n (% of total) | Downstaged | True T2N0 | Upstaged | p | |
|---|---|---|---|---|---|---|
| Tumor Characteristics | ||||||
| Grade | <0.0001 | |||||
| 1 | 94 (6.2%) | 32 (34.0%) | 47 (50.0%) | 15 (16.0%) | ||
| 2 | 644 (42.4%) | 153 (23.8%) | 243 (37.8%) | 248 (38.5%) | ||
| 3 | 638 (42.0%) | 83 (13.1%) | 209 (32.8%) |
|
||
| 4 | 27 (1.8%) | 5 (18.5%) | 7 (25.9%) | |||
| Missing | 117 (7.7%) | 37 (31.6%) | 43 (36.7%) | 37 (31.6%) | ||
| Histology | 0.0004 | |||||
| Squamous | 324 (21.3%) | 55 (17.0%) | 137 (42.3%) | 132 (40.7%) | ||
| Adenocarcinoma | 1088 (71.6%) | 245 (22.5%) | 376 (34.6%) | 467 (42.9%) | ||
| Other | 108 (7.1%) | 10 (9.2%) | 36 (33.3%) | 62 (57.4%) | ||
| Lymphovascular Invasion | <0.0001 | |||||
| Present | 92 (6.1%) | 2 (2.2%) | 17 (18.5%) |
|
||
| Absent | 137 (9.0%) | 12 (8.8%) | 61 (44.5%) | 64 (46.7%) | ||
| Missing | 1291 (85.0%) | 296 (22.9%) | 471 (36.5%) | 524 (40.6%) | ||
| Tumor Size | <0.0001 | |||||
| Missing | 259 (17.0%) | 68 (26.2%) | 104 (40.2%) | 87 (33.6%) | ||
| <1cm | 74 (4.9%) | 29 (39.2%) | 33 (44.6%) | 12 (16.2%) | ||
| 1 to <2cm | 241 (15.9%) | 74 (30.7%) | 88 (36.5%) | 79 (32.8%) | ||
| 2 to <3cm | 329 (21.6%) | 59 (17.9%) | 118 (35.6%) | 152 (46.2%) | ||
| 3 to <4cm | 250 (16.4%) | 42(16.8%) | 85 (34.0%) |
|
||
| 4 to <5 cm | 267 (17.6%) | 20 (12.0%) | 50 (29.9%) | |||
| 5cm + | 200 (13.2%) | 18 (9.0%) | 71 (35.5%) | |||
| Patient Characteristics | ||||||
| Age | 0.5849 | |||||
| mean +/− SD | 64.4 +/− 10.0 | 65.1 +/− 11.1 | 64.5 +/− 11.0 | |||
| median (IQR) | 65.5 (58–72) | 66 (57–74) | 65 (58–73) | |||
| Sex | 0.0078 | |||||
| Male | 1234 (81.2%) | 264 (21.4%) | 424 (34.4%) | 546 (44.3%) | ||
| Female | 286 (18.8%) | 46 (16.1%) | 125 (43.7%) | 115 (40.2%) | ||
| Race | 0.8468 | |||||
| White | 1393 (91.6%) | 287 (20.6%) | 500 (35.9%) | 606 (43.5%) | ||
| Black | 77 (5.1%) | 12 (15.6%) | 31 (40.3%) | 34 (44.2%) | ||
| Other | 50 (3.3%) | 11 (22.0%) | 18 (36.0%) | 21 (42.0%) | ||
| Urban/Rural County | 0.1104 | |||||
| Metropolitan | 1169 (76.9%) | 242 (20.7%) | 424 (36.7%) | 503 (43.0%) | ||
| Urban | 230 (15.1%) | 34 (14.8%) | 84 (36.5%) | 112 (48.7%) | ||
| Rural | 30 (2.0%) | 10 (33.3%) | 9 (30.0%) | 11 (36.7%) | ||
| Cancer Center Type | 0.0007 | |||||
| Community | 674 (44.3%) | 112 (16.6%) | 272 (40.4%) | 290 (43.0%) | ||
| Academic | 846 (55.7%) | 198 (23.4%) | 277 (32.7%) | 371 (43.9%) | ||
Of the features that were statistically significant, tumor grade, presence of lymphovascular invasion, and tumor size all had clinically meaningful categories that placed a patient above the threshold of 48.1% probability of upstaging, where the patients may be expected to benefit from routine induction chemoradiation. These categories are shown in Table 3. For patients with tumor grade 3 and 4, 54.2% and 55.6% respectively were ultimately pathologically upstaged. Patients with evidence of lymphovascular invasion were upstaged 79.4% of the time. Tumor size was also associated with upstaging: for tumors 3–4 cm, the rate of upstaging was 49.2%; for tumors 4–5 cm, the rate of upstaging was 58.1%; and for tumors greater than 5 cm, the rate of upstaging was 55.5%. Tumor size greater than 4 cm and presence of lymphovascular invasion placed the patients above the threshold for benefit of induction chemoradiation, even when perioperative mortality for that strategy was doubled within our model. Univariable and multivariable analyses confirmed these variables were significantly associated with upstaging (all p<0.0001) versus same- or downstaging, and the results of these analyses and corresponding odds ratios can be seen in Table 4.
Table 4.
Univariable and multivariable analysis of selected covariates with likelihood of upstaging as the dependent variable.
| Variable | Univariable Analysis |
p | Multivariable Analysis |
p | |
|---|---|---|---|---|---|
| Odds Ratio (95% CI) | Odds Ratio (95% CI) | ||||
| Tumor Characteristics | |||||
| Grade | <.0001 | <.0001 | |||
| 1 | Reference | Reference | |||
| 2 | 3.3 (1.9–5.9) | 2.6 (1.3–5.1) | |||
| 3 | 6.2 (3.5–11.1) | 5.6 (2.9–10.8) | |||
| 4 | 6.6 (2.6–16.8) | 3.8 (1.4–10.9) | |||
| Missing | 2.4 (1.2–4.8) | 4.2 (1.9–9.5) | |||
| Histology | 0.009 | ||||
| Squamous | 0.9 (0.7–1.2) | ||||
| Adenocarcinoma | Reference | ||||
| Other | 1.7 (1.2–2.7) | ||||
| Lymphovascular Invasion | <.0001 | <.0001 | |||
| Present | 5.6 (3.4–9.4) | 4.6 (2.6–7.9) | |||
| Absent | 1.3 (0.9–1.8) | 1.4 (1.0–2.1) | |||
| Missing | Reference | Reference | |||
| Tumor Size | <.0001 | <.0001 | |||
| <1cm | Reference | Reference | |||
| 1 to <2cm | 2.5 (1.3–4.9) | 3.1 (1.4–6.6) | |||
| 2 to <3cm | 4.4 (2.3–8.5) | 5.3 (2.5–11.0) | |||
| 3 to <4cm | 5.0 (2.6–9.7) | 5.5 (2.6–11.6) | |||
| 4 to <5 cm | 7.2 (3.6–14.3) | 7.9 (3.6–17.0) | |||
| 5cm + | 6.4 (3.3–12.7) | 7.3 (3.4–15.7) | |||
| Patient Characteristics | |||||
| Age | 1.0 (0.99–1.01) | 0.531 | |||
| Sex: Male vs Female | 1.2 (0.9–1.5) | 0.215 | |||
| Race: White | Reference | 0.971 | |||
| Black | 1.0 (0.6–1.6) | ||||
| Other | 0.9 (0.5–1.7) | ||||
| Urban/Rural County: Metropolitan | Reference | 0.209 | Reference | 0.069 | |
| Urban | 1.3 (0.9–1.7) | 1.3 (0.9–1.8) | |||
| Rural | 0.8 (0.4–1.6) | 0.4 (0.2–1.1) | |||
| Cancer Center: Academic vs Community | 1.0 (0.8–1.2) | 0.747 | |||
DISCUSSION
Our decision analysis allows us to integrate the best available clinical evidence for the role of induction therapy in cT2N0 patients. Our baseline model demonstrated that, overall, patients with cT2N0 cancer who underwent upfront esophagectomy have an estimated survival benefit of approximately 2–3 months compared to patients who underwent induction therapy followed by esophagectomy. This benefit is likely due to the effect of the true lower stage patients in our model. Survival for patients with less than pT2N0 is better with upfront surgery4 due to avoidance of unnecessary and likely harmful treatment in this population that could cause deconditioning, chemoradiation-related morbidity or mortality, treatment delay, and potentially increased surgical complications and mortality. This result demonstrates that identification of higher risk patients for induction therapy would be expected to yield better results than taking a uniform approach to this group.
Our findings also show that the theoretical threshold of EUS understaging (or pathological upstaging) at which induction chemoradiation becomes beneficial is within the published range of diagnostic uncertainty. The results from this decision analysis can be used to guide clinical recommendations for a patient with cT2N0 cancer based on whether the patient’s probability of pathological upstaging is greater than 48.1%. Our analysis of tumor factors from the NCDB identified several high-risk features that should raise suspicion for a true higher pathological stage. It would be reasonable to routinely use induction chemoradiation for patients with any one of the following features placing a patient above the threshold for benefit: tumor grade 3 or 4, presence of lymphovascular invasion, or tumor size greater than 3 cm. Tumor size can be estimated from an endoscopic exam and grade can often be assessed from a routine biopsy. Identification of lymphovascular invasion may depend on the size and quality of a biopsy in addition to pathologic staining techniques,16,22,23 but can occasionally be determined preoperatively24,25 and, when available, may be very helpful in guiding treatment.
Other universally available clinical features can also be considered in the application of our model: specifically, dysphagia and PET SUV. While no studies have explicitly examined the probability of upstaging in cT2N0 patients with dysphagia, several series on resectable esophageal cancer have shown the presence of dysphagia to have a specificity for “at least T3 disease” of greater than 88%.26–28 Similarly, studies examining PET SUV have not been done exclusively in cT2N0 patients, however a SUVmax greater than 2.5 in the primary tumor has been associated with presence of nodal disease.29,30 Presence of these factors will likely place a cT2N0 patient above the 48.1% threshold for benefit of induction chemoradiation and should be considered during clinical decision-making about the best strategy.
This study has some limitations that should be considered. First, data for the model was derived from existing published studies and relied, at least partially, on retrospective reviews which may be affected by selection bias. Patients who received adjuvant therapy likely had a smooth operative course and were healthy enough to tolerate this treatment. There may also be surgeon bias in selecting low-risk patients for upfront surgery and high-risk patients for induction therapy which may affect survival estimates for the pathologically staged subgroups. We did utilize data from randomized controlled trials where available. However, these data may also be biased by the performance status needed to enroll in such trials: progression to inoperability after induction therapy as well as the average time interval from neoadjuvant treatment to surgery may be underestimated when considering the frail patient with esophageal cancer. Second, for some survival data points we relied on cohorts that were quite similar to, but not exact matches for those in our model. One example is the use of survival data for planned definitive chemoradiation patients as the expected survival for patients unable to undergo an operation after induction therapy. However, the same factors that made these patients non-operative candidates would likely result in shortened expected overall survival compared to those with fewer comorbidities, and therefore we felt this group was appropriately representative. Additionally, other papers have reported proportionally higher percentages of chemoradiation-related mortality, sometimes up to 4%.31 These proportions occurred in studies with smaller sample sizes and the absolute number of deaths remained low, therefore, we retained the lower estimated prevalence of death related to chemoradiation. Finally, all patients in the NCDB may not have received the CROSS regimen of induction therapy which has low morbidity and mortality, and this may affect survival estimates. When considering these limitations, though, it is worth noting that in development and testing of our decision tree, none of these individual factors were primary drivers of the model outcome.
We feel these limitations are significantly outweighed by several major strengths of our approach. We incorporated the best-available existing evidence into a model that can inform clinical practice, with both the information provided from the baseline model as well as through our sensitivity analyses. We supplemented published data with primarily derived outcomes for specific pathologic subgroups using a large, validated dataset, obtaining survival information reflecting real-world, practice-based outcomes. Through sensitivity analyses, we were able to identify the drivers of our model and the threshold at which each treatment paradigm will confer improved survival. We analyzed clinical and pathologic features that would identify patients at high risk of upstaging, and consideration of these features in the context of our sensitivity analysis (Figure 2) would allow for estimation of the expected survival benefit of induction chemoradiation for an individual. This may be helpful for providers in making treatment recommendations and risk-benefit discussions with patients. Additionally, our model implicitly accounts for patients who had a margin positive resection and postoperative complications – both of which would be expected to confer a worse overall survival with or without induction therapy. Rates for these outcomes were within similar ranges across publications used for our model input, and the estimated median survival incorporates patients with these poor prognostic factors. These features of our model allow for broad applicability.
In conclusion, this decision analysis provides evidence that induction chemoradiation improves survival in patients with cT2N0 esophageal cancer if the probability of EUS understaging is greater than 48.1%. Upfront surgery is preferred when the error rate of EUS is below this threshold. Several clinical factors including high tumor grade, presence of lymphovascular invasion, large tumor size, presence of dysphagia, or PET SUV greater than 2.5 increase the probability of EUS understaging and consequent pathologic upstaging and can guide a clinician to consider induction therapy. While additional patient-specific factors must be considered with any clinical decision-making, this study helps to provide guidance in the initial management of cT2N0 in face of inherent diagnostic uncertainty of EUS staging.
Table 5.
Summary of conclusions and supporting results.
| Analysis | Conclusion | Supporting Results |
|---|---|---|
| Baseline Decision Analysis Model | Survival is similar between upfront surgery and induction chemoradiation, with a small expected survival benefit of 2.5 months for upfront surgery. | Upfront esophagectomy yields an expected overall survival of 48.3 months compared to 45.9 months for induction chemoradiation. |
| Sensitivity Analysis of Diagnostic Uncertainty | Induction chemoradiation is expected to yield a survival advantage within the published range of EUS understaging for cT2N0 disease: 32–65%. | The threshold for benefit is with a probability of upstaging of >48.1% (see Figure 2). |
| Descriptive Analysis of NCDB Patient Characteristics by True Stage | Patients with a tumor grade of 3 or 4, tumor size >3cm, or presence of lymphovascular invasion would be expected to gain a survival benefit from induction chemoradiation. | With any one of these features, the proportion of patients that were upstaged was above the threshold for benefit of 48.1% (see Table 3). |
| Univariable and Multivariable Analysis of NCDB Patient Characteristics | Tumor grade, tumor size, and presence of lymphovascular invasion were found to be associated with risk of pathological upstaging. | These features were significant on univariable and multivariable analysis, all p<0.0001 (see Table 4). |
Central Picture Legend.
Induction chemoradiation is beneficial with a 48.1% or higher probability of upstaging.
Central Message.
Induction chemoradiation may offer a survival benefit over upfront surgery for clinical T2N0 esophageal cancer patients who have a greater than 48% probability of pathologic upstaging.
Perspective Statement.
This research aims to address the controversy in treating cT2N0 esophageal cancer that exists because of the inaccuracies in endoscopic ultrasound staging. We created a decision analysis model to compare two treatment paradigms for cT2N0 patients: 1) upfront esophagectomy and 2) induction chemoradiation followed by esophagectomy, focusing on the impact of diagnostic inaccuracy on patient outcomes.
Acknowledgments
Funding: TRS was supported by National Institutes of Health (NIH) grant number 2T32HL7776-21. RZP and LCE were supported by NIH 5T32CA00962128. JLH was supported by NIH 5T32CA009621-27. S-HC was supported by the Agency for Healthcare Research and Quality (AHRQ) grant K01 HS022330 and NIH R21 DK110530.VP was supported by K07CA178120.
Glossary of Abbreviations
- cT2N0
clinically staged T2N0
- pT2N0
pathologically staged T2N0
- ypT2N0
post-induction pathologically staged T2N0
- EUS
endoscopic ultrasound
- T
tumor stage
- N
nodal stage
- NCDB
National Cancer Database
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflicts of Interest: none for any of the authors
IRB Approval: Exempt.
References
- 1.National Cancer Institute. Surveillance, Epidemiology and ERP. [Accessed June 19, 2017];Cancer of the Esophagus - Cancer Stat Facts. Cancer Statistics. https://seer.cancer.gov/statfacts/html/esoph.html. Published 2017.
- 2.Rice TW, Blackstone EH, Rusch VW. 7th Edition of the AJCC Cancer Staging Manual: Esophagus and Esophagogastric Junction. Ann Surg Oncol. 2010;17(7):1721–1724. doi: 10.1245/s10434-010-1024-1. [DOI] [PubMed] [Google Scholar]
- 3.Mariette C, Dahan L, Mornex F, et al. Surgery alone versus chemoradiotherapy followed by surgery for stage I and II esophageal cancer: final analysis of randomized controlled phase III trial FFCD 9901. J Clin Oncol. 2014;32(23):2416–2422. doi: 10.1200/JCO.2013.53.6532. [DOI] [PubMed] [Google Scholar]
- 4.Rice TW, Mason DP, Murthy SC, et al. T2N0M0 esophageal cancer. J Thorac Cardiovasc Surg. 2007;133(2):317–324. doi: 10.1016/j.jtcvs.2006.09.023. [DOI] [PubMed] [Google Scholar]
- 5.van Hagen P, Hulshof MCCM, van Lanschot JJB, et al. Preoperative Chemoradiotherapy for Esophageal or Junctional Cancer. N Engl J Med. 2012;366(22):2074–2084. doi: 10.1056/NEJMoa1112088. [DOI] [PubMed] [Google Scholar]
- 6.Chen W-H, Chao Y-K, Chang H-K, et al. Long-term outcomes following neoadjuvant chemoradiotherapy in patients with clinical T2N0 esophageal squamous cell carcinoma. Dis esophagus Off J Int Soc Dis Esophagus. 2012;25(3):250–255. doi: 10.1111/j.1442-2050.2011.01243.x. [DOI] [PubMed] [Google Scholar]
- 7.Dolan JP, Kaur T, Diggs BS, et al. Significant understaging is seen in clinically staged T2N0 esophageal cancer patients undergoing esophagectomy. Dis esophagus Off J Int Soc Dis Esophagus. 2016;29(4):320–325. doi: 10.1111/dote.12334. [DOI] [PubMed] [Google Scholar]
- 8.Speicher PJ, Ganapathi AM, Englum BR, et al. Induction Therapy Does Not Improve Survival for Clinical Stage T2N0 Esophageal Cancer. J Thorac Oncol. 2014;9(8):1195–1201. doi: 10.1097/JTO.0000000000000228. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zhang JQ, Hooker CM, Brock MV, et al. Neoadjuvant chemoradiation therapy is beneficial for clinical stage T2 N0 esophageal cancer patients due to inaccurate preoperative staging. Ann Thorac Surg. 2012;93(2) doi: 10.1016/j.athoracsur.2011.10.061. 429-35-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Crabtree TD, Yacoub WN, Puri V, et al. Endoscopic ultrasound for early stage esophageal adenocarcinoma: implications for staging and survival. Ann Thorac Surg. 2011;91(5) doi: 10.1016/j.athoracsur.2011.01.063. 1509-15-6. [DOI] [PubMed] [Google Scholar]
- 11.Stiles BM, Mirza F, Coppolino A, et al. Clinical T2-T3N0M0 esophageal cancer: the risk of node positive disease. Ann Thorac Surg. 2011;92(2) doi: 10.1016/j.athoracsur.2011.04.004. 491-6-8. [DOI] [PubMed] [Google Scholar]
- 12.Crabtree TD, Kosinski AS, Puri V, et al. Evaluation of the reliability of clinical staging of T2 N0 esophageal cancer: a review of the Society of Thoracic Surgeons database. Ann Thorac Surg. 2013;96(2):382–390. doi: 10.1016/j.athoracsur.2013.03.093. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hofstetter W. Treatment of clinical T2N0M0 esophageal cancer. Ann Surg Oncol. 2014;21(12):3713–3714. doi: 10.1245/s10434-014-3935-8. [DOI] [PubMed] [Google Scholar]
- 14.Hardacker TJ, Ceppa D, Okereke I, et al. Treatment of Clinical T2N0M0 Esophageal Cancer. Ann Surg Oncol. 2014;21(12):3739–3743. doi: 10.1245/s10434-014-3929-6. [DOI] [PubMed] [Google Scholar]
- 15.Tekola BD, Sauer BG, Wang AY, White GE, Shami VM. Accuracy of endoscopic ultrasound in the diagnosis of T2N0 esophageal cancer. J Gastrointest Cancer. 2014;45(3):342–346. doi: 10.1007/s12029-014-9616-9. [DOI] [PubMed] [Google Scholar]
- 16.Samson P, Puri V, Robinson C, et al. Clinical T2N0 Esophageal Cancer: Identifying Pretreatment Characteristics Associated With Pathologic Upstaging and the Potential Role for Induction Therapy. Ann Thorac Surg. 2016;101(6):2102–2111. doi: 10.1016/j.athoracsur.2016.01.033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.van Hagen P, Hulshof MCCM, van Lanschot JJB, et al. Preoperative chemoradiotherapy for esophageal or junctional cancer. N Engl J Med. 2012;366(22):2074–2084. doi: 10.1056/NEJMoa1112088. [DOI] [PubMed] [Google Scholar]
- 18.Abrams JA, Buono DL, Strauss J, McBride RB, Hershman DL, Neugut AI. Esophagectomy compared with chemoradiation for early stage esophageal cancer in the elderly. Cancer. 2009;115(21):4924–4933. doi: 10.1002/cncr.24536. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lapar DJ, Stukenborg GJ, Lau CL, Jones DR, Kozower BD. Differences in reported esophageal cancer resection outcomes between national clinical and administrative databases. J Thorac Cardiovasc Surg. 2012;144(5):1152–1157. doi: 10.1016/j.jtcvs.2012.08.010. [DOI] [PubMed] [Google Scholar]
- 20.Medical Research Council Oesophageal Cancer Working Group. Surgical resection with or without preoperative chemotherapy in oesophageal cancer: a randomised controlled trial. Lancet. 2002;359(9319):1727–1733. doi: 10.1016/S0140-6736(02)08651-8. [DOI] [PubMed] [Google Scholar]
- 21.Fiorica F, Di Bona D, Schepis F, et al. Preoperative chemoradiotherapy for oesophageal cancer: a systematic review and meta-analysis. [Accessed December 6, 2017];Gut. 2004 53(7):925–930. doi: 10.1136/gut.2003.025080. http://www.ncbi.nlm.nih.gov/pubmed/15194636. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Saad RS, Lindner JL, Liu Y, Silverman JF. Lymphatic Vessel Density as Prognostic Marker in Esophageal Adenocarcinoma. Am J Clin Pathol. 2009;131(1):92–98. doi: 10.1309/AJCPKWUQSIPVG90H. [DOI] [PubMed] [Google Scholar]
- 23.Goscinski MA, Nesland JM, Giercksky K-E, Dhakal HP. Primary tumor vascularity in esophagus cancer. CD34 and HIF1-α expression correlate with tumor progression. Histol Histopathol. 2013;28(10):1361–1368. doi: 10.14670/HH-28.1361. [DOI] [PubMed] [Google Scholar]
- 24.Cen P, Hofstetter WL, Lee JH, et al. Value of endoscopic ultrasound staging in conjunction with the evaluation of lymphovascular invasion in identifying low-risk esophageal carcinoma. Cancer. 2008;112(3):503–510. doi: 10.1002/cncr.23217. [DOI] [PubMed] [Google Scholar]
- 25.Cen P, Hofstetter WL, Correa AM, et al. Lymphovascular invasion as a tool to further subclassify T1b esophageal adenocarcinoma. Cancer. 2008;112(5):1020–1027. doi: 10.1002/cncr.23265. [DOI] [PubMed] [Google Scholar]
- 26.Fang TC, Oh YS, Szabo A, Khan A, Dua KS. Utility of dysphagia grade in predicting endoscopic ultrasound T-stage of non-metastatic esophageal cancer. Dis Esophagus. 2016;29(6):642–648. doi: 10.1111/dote.12394. [DOI] [PubMed] [Google Scholar]
- 27.Ripley RT, Sarkaria IS, Grosser R, et al. Pretreatment Dysphagia in Esophageal Cancer Patients May Eliminate the Need for Staging by Endoscopic Ultrasonography. Ann Thorac Surg. 2016;101(1):226–230. doi: 10.1016/j.athoracsur.2015.06.062. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Schlansky B, Dimarino AJ, Loren D, Infantolino A, Kowalski T, Cohen S. A survey of oesophageal cancer: pathology, stage and clinical presentation. Aliment Pharmacol Ther. 2006;23(5):587–593. doi: 10.1111/j.1365-2036.2006.02782.x. [DOI] [PubMed] [Google Scholar]
- 29.I HS, Kim S-J, Kim IJ, Kim K. Predictive Value of Metabolic Tumor Volume Measured by 18F-FDG PET for Regional Lymph Node Status in Patients With Esophageal Cancer. Clin Nucl Med. 2012;37(5):442–446. doi: 10.1097/RLU.0b013e318238f703. [DOI] [PubMed] [Google Scholar]
- 30.Shin S, Kim HK, Choi YS, Kim K, Shim YM. Clinical stage T1–T2N0M0 oesophageal cancer: accuracy of clinical staging and predictive factors for lymph node metastasis†. Eur J Cardio-Thoracic Surg. 2014;46(2):274–279. doi: 10.1093/ejcts/ezt607. [DOI] [PubMed] [Google Scholar]
- 31.Tepper J, Krasna MJ, Niedzwiecki D, et al. Phase III trial of trimodality therapy with cisplatin, fluorouracil, radiotherapy, and surgery compared with surgery alone for esophageal cancer: CALGB 9781. J Clin Oncol. 2008;26(7):1086–1092. doi: 10.1200/JCO.2007.12.9593. [DOI] [PMC free article] [PubMed] [Google Scholar]