Abstract
Background:
The poor prognosis of esophageal adenocarcinoma (EAC) has focused efforts on early detection by serial endoscopic surveillance of Barrett’s esophagus (BE). Previously, we reported that receipt of endoscopy before EAC diagnosis was associated with improved survival.
Aim:
We aimed to refine our previous analysis, assessing surveillance as measured by performance of serial endoscopy before EAC diagnosis and evaluating its association with stage and survival.
Methods:
A retrospective cohort study was performed using the Surveillance, Epidemiology and End Results (SEER)-Medicare database. Patients aged ≥ 70 years with EAC diagnosed 1998–2009 were identified. Diagnosis with BE and receipt of ≥ 2 upper endoscopic procedures within 5 years before cancer diagnosis were identified. We compared a reference group not receiving serial endoscopy to 3 patterns based on ≥ 2 endoscopy dates relative to a timepoint 2 years before cancer diagnosis: “remote,” “recent,” and “sustained.”
Results:
Among 5,532 patients, 28% (n=1,575) had localized stage. Thirteen percent (n=703) received ≥ 2 endoscopic procedures before cancer diagnosis: 224, 298, and 181 in the “recent,” “remote,” and “sustained” groups. Serial endoscopy and prior BE were associated with localized stage (“sustained” group OR 2.95, 95% confidence interval [CI] 2.07, 4.19; prior BE OR 2.68, 95% CI 2.03, 3.56). Serial endoscopy was associated with improved survival even with adjustment for lead time bias (“sustained” group HR 0.45, 95% CI 0.37, 0.55) and length time bias.
Conclusions:
Sustained endoscopy was associated with earlier stage and improved survival. These results support the role of sustained surveillance in early detection of EAC.
Keywords: esophageal neoplasm, endoscopy, stage, survival
Introduction
An estimated 19,260 new cases of esophageal cancer will be diagnosed in the USA in 2021, with 15,530 deaths[1]. In the USA and much of the Western world, the predominant histologic subtype is adenocarcinoma. Despite advances in treatment, the overall 5-year survival rate in the USA is approximately 17% [2], reflecting this cancer’s propensity to present with lymph node involvement or distant spread. Risk factors for esophageal adenocarcinoma (EAC) include central obesity, gastroesophageal reflux disease (GERD), male gender, and white race [3]. The overall poor prognosis of esophageal adenocarcinoma highlights the need for early detection. However, per US gastrointestinal society guidelines [4–6], screening for Barrett’s Esophagus (BE), the precursor lesion for EAC, should only be considered in select patients with multiple risk factors for EAC/BE.
Surveillance of nondysplastic BE is a conditional recommendation in the 2019 American Society for Gastrointestinal Endoscopy guidelines [6] since there is conflicting evidence regarding the effectiveness of BE surveillance. A recent systematic review and meta-analysis found that surveillance was associated with detection of EAC at earlier stages [7]. While some studies have reported that surveillance of BE was associated with improved mortality from EAC [8, 9], other studies have revealed no mortality benefit with BE surveillance [10] or with BE diagnosis prior to development of EAC [11]. Due to the low incidence of EAC in BE, prospective studies evaluating this issue are hampered by the need for a large sample size. While a randomized trial is currently underway (BOSS trial, NCT00987857) [12], results are not yet available.
In a prior study using a population-based database, we found that receipt of upper endoscopy prior to EAC diagnosis was associated with improved survival [13]. We sought to refine our analysis by assessing use of serial endoscopy prior to EAC diagnosis in a U.S. population. We aimed to evaluate the association of serial (i.e., ≥ 2) endoscopic procedures prior to cancer diagnosis with stage at presentation and overall survival. We hypothesized that timing of serial endoscopic procedures within a 5-year period preceding EAC diagnosis might impact stage at presentation and overall survival.
Methods
This retrospective cohort study used the Surveillance, Epidemiology & End Results (SEER) tumor registry linked to Medicare claims data. Patients diagnosed with esophageal adenocarcinoma were identified from SEER, and endoscopic procedures prior to cancer diagnosis were identified from Medicare claims. The study was approved by the University Hospitals Cleveland Medical Center (UHCMC) Institutional Review Board (IRB) and the National Cancer Institute (NCI).
SEER-Medicare Database
The SEER program of the National Cancer Institute collects data on cancer incidence and survival in the United States in specific population-based registries currently covering approximately 34% of the US population [14]. Individual identifiers in the SEER files are matched with identifiers in Medicare’s master enrollment file, allowing identification of comorbid conditions and patient encounters for individuals aged ≥ 65 years diagnosed with cancer in SEER geographic regions. Medicare claims files include hospital inpatient claims, physician-supplier claims, or outpatient claims. Procedures can be identified from physician-supplier or outpatient claims files using Current Procedural Terminology codes, while inpatient claims contained International Classification of Diseases, 9th Revision-Clinical Modification (ICD-9-CM) procedure codes prior to October 2016.
Measures
Study Population
Patients aged ≥ 70 years with incident esophageal adenocarcinoma diagnosed 1998–2009 were identified from the SEER Patient Entitlement and Diagnosis Summary File (PEDSF) based on anatomical site (restricted to esophagus) and ICD-O-3 histology codes for adenocarcinoma [15]. Age 70 was chosen as a minimum age cutoff to allow for identification of endoscopic procedures starting 5 years prior to cancer diagnosis. Patients enrolled in a Health Maintenance Organization or not enrolled in Medicare Part B from 5 years before to 3 months after cancer diagnosis were excluded due to incomplete claims data. The SEER staging system classifies esophageal cancer stage at presentation as carcinoma in situ, localized, regional, or distant. Cases with carcinoma in situ were excluded because the study focused on invasive cancers, and carcinoma in situ may be incompletely captured in SEER. Patient demographic characteristics (age, race, gender, SEER registry) were obtained from the PEDSF. County level data on income and education level, obtained by linking Federal Information Processing Standard codes to the Area Health Resources File, were used as proxies for socioeconomic status. Area Health Resources File data on income and education level were categorized into quartiles due to their skewed distribution. A Charlson comorbidity score was calculated using ICD-9-CM diagnosis codes [16, 17] from inpatient, outpatient, and physician-supplier claims in the period 12 months prior to cancer diagnosis. The Charlson score is a weighted index measure of conditions including cardiopulmonary diseases, liver disease, diabetes mellitus, chronic kidney disease, hemiplegia, and cerebrovascular accident that may impact mortality. For the current study, malignancy was excluded from the Charlson score.
Assessment of Endoscopy and Barrett’s Esophagus
Use of endoscopy was identified from procedure codes CPT 43234, 43235, 43239, and 43255 and ICD-9-CM 45.13, 45.14, and 45.16. Patients undergoing serial endoscopy were defined as individuals who received ≥ 2 endoscopies during the time period from 5 years to 3 months prior to cancer diagnosis, hereafter referred to as the “prediagnosis” period. The presence of an ICD-9-CM diagnosis code for Barrett’s esophagus (530.2 or 530.85) [18] was assessed in this same time period. Among patients receiving serial endoscopy, the interval between the last endoscopic procedure during the prediagnosis period and cancer diagnosis was determined. Due to the skewed distribution of this interval, 3 endoscopic patterns were defined (Fig. 1) based on timing of endoscopic procedures relative to a timepoint 2 years prior to cancer diagnosis: “remote,” “recent,” and “sustained” groups. These serial endoscopy groups were compared to a reference group of patients not receiving ≥ 2 endoscopies in the prediagnosis period.
Figure 1.
Classification scheme for serial endoscopy in the period 5 years to 3 months prior to esophageal adenocarcinoma diagnosis.
Assessment of Esophagectomy, Radiation, and Chemotherapy
Due to its potential impact on overall survival, esophagectomy was identified from MedPAR inpatient files (ICD-9-CM codes 42.40–42.42, 43.99, 42.50–42.59, 42.60–42.69, 43.5, 43.99). As in a prior study, radiation therapy was identified from ICD-9 codes 92.2, V58.0, V66.1, and V67.1 and CPT code 77xxx. Chemotherapy was identified from the following codes: administration of chemotherapy (99.25), antineoplastic chemotherapy encounter (V58.11), cancer chemotherapy follow-up (V67.2), convalescence and palliative care after chemotherapy (V66.2), chemotherapy administration (Q0083-Q0085), chemotherapeutic agents (Healthcare Common Procedure Coding System codes J9000-J9999), and revenue center codes (0331, 0332, 0335) [19].
Statistical Analysis
Baseline characteristics by presenting stage (localized vs. other) were compared using descriptive statistics. Logistic regression models were developed to identify factors associated with localized stage at presentation. Overall survival was measured from date of diagnosis until death. Survival times were administratively censored at end of follow-up December 31, 2013. Kaplan-Meier curves were generated to compare overall survival by serial endoscopic pattern. Cox proportional hazards models were developed to identify factors associated with time to death, focusing on the impact of a prior BE diagnosis and serial endoscopic pattern.
Lead Time Bias Adjustment
Adjustment for lead time bias was performed. Lead time bias refers to improved survival from earlier diagnosis due to screening compared with diagnosis from clinically symptomatic disease [20]. Lead time therefore applies to screen-detectable cancers rather than cancers diagnosed due to worrisome symptoms. We subdivided the cohort into 2 subgroups (Fig. 2): 1) Symptomatic (SYM) subgroup, based on presence of diagnosis codes for concerning symptoms suggestive of malignancy (dysphagia, anemia, blood loss, gastrointestinal bleeding, weight loss, or abnormal imaging) within 1 month prior to or after cancer diagnosis; and 2) Screening (SCR) subgroup, based on the absence of diagnosis codes for these concerning symptoms (Supplementary Table 1). Symptoms suggestive of malignancy were similar to those in a previous BE study utilizing VA data [9]. Since patients in the SYM subgroup had symptoms suggestive of malignancy, no adjustment was made to their survival times for lead time bias. Survival times for patients in the SCR subgroup were adjusted for lead time bias using the previous method published by Duffy et al [20]. Duffy’s method assumes that the sojourn time – the time period between which a cancer is detectable by screening and a cancer is detectable by symptoms – has an exponential distribution. Based on prior literature [7, 9, 11, 21], we assumed that the median sojourn time was 2.5 years, with sensitivity analyses performed using 2 and 3 years.
Figure 2.
Approach to lead time bias adjustment. Dx, Diagnosis. SCR, Screening subgroup. SYM, Symptomatic subgroup.
1 Dysphagia, anemia/blood loss, gastrointestinal bleeding, weight loss, abnormal imaging
Length Time Bias Adjustment
The method of Duffy [20] was also used to examine the possible impact of length bias, using data already adjusted for lead time bias, focusing on risk ratios for 1.5-year death rates comparing the “sustained,” “remote,” and “recent” groups to the reference group. It is assumed there are two latent types of tumors, more aggressive (A) and less aggressive (B), in proportions q and (1 − q), respectively, where the ratio of death rates of aggressive to non-aggressive is θ (θ ≤ 1) and those with type B tumors are 1/θ times more likely to be asymptomatic at diagnosis, which induces length bias in the observed risk ratio of death at 1.5 years. In a sensitivity analysis, the true risk ratio φ comparing death rates of asymptomatic versus symptomatic patients at diagnosis is then calculated over a range of values for the unknown values q and θ suggested and used by others [11, 20, 21]. Values of the estimated true risk ratios were then used to obtain length-bias adjusted estimates of overall 1.5-year survival in the four groups, and relative risks comparing the three non-reference groups to the reference group under conservative assumptions (favoring a larger length bias effect). See Supplementary Methods for more details.
SAS version 9.3 (Cary, NC) was used for all analyses.
Results
A flowsheet displaying stepwise selection of the cohort is shown (Fig. 3). After inclusion and exclusion criteria, 5,532 patients with incident esophageal adenocarcinoma were identified. Among these, 28.4% (n=1,575), 25.9% (n=1,437), and 29.3% (n=1,625) presented with localized, regional, and distant SEER stage, respectively, while 16.1% (n=895) were unstaged. As a result, 71.5% (n=3,957) presented with a stage other than localized disease. The mean follow-up time in the overall cohort was 1.22 years (median, 0.56 years).
Figure 3.
Stepwise selection of cohort. The proportion of patients excluded at each step is provided in parentheses. HMO, Health Maintenance Organization. SCC, squamous cell carcinoma. CIS, carcinoma in situ.
Baseline characteristics by stage at presentation (localized versus other stage) are shown in Table 1. Overall, 6% (n=329) had a prior diagnosis of Barrett’s esophagus. However, patients presenting with localized stage were more likely to have a prior diagnosis of Barrett’s esophagus (14%, n=213 vs. 3%, n=116, p < 0.0001). The overall esophagectomy rate was 18% (n=991). Patients presenting with localized stage were more likely to undergo esophagectomy (26%, n=407 vs. 15%, n=584, p < 0.0001). Receipt of serial endoscopy was more common in patients presenting with localized stage: 8% (n=128), 8% (n=124), and 7% (n=115) of patients in the localized stage group were in the “remote,” “recent,” and “sustained” serial endoscopy groups, respectively, compared with 4% (n=170), 3% (n=100), and 2% (n=66) of patients presenting with other stage (p<0.0001). Factors associated with localized stage at presentation are presented in Table 2. Serial endoscopy and prior BE were associated with localized stage (“sustained” group odds ratio [OR] 2.95, 95% confidence interval [CI] 2.07, 4.19; prior BE OR 2.68, 95% CI 2.03, 3.56).
Table 1.
Baseline characteristics by stage at presentation (localized versus other stage)
| Localized (n=1,575) |
Other Stage (n=3,957) |
P value | |
|---|---|---|---|
| Mean age (SD) | 79 (6) | 79 (6) | 0.61 |
| Female gender | 324 (21) | 865 (22) | 0.29 |
| Black/Other race | 72 (5) | 202 (5) | 0.41 |
| Charlson score | |||
| 0 | 346 (22) | 1,110 (28) | <0.0001 |
| 1 | 1,078 (68) | 2,519 (64) | |
| ≥ 2 | 151 (10) | 328 (8) | |
| Prior Barrett’s esophagus diagnosis | 213 (14) | 116 (3) | <0.0001 |
| Surgery | 407 (26) | 584 (15) | <0.0001 |
| Radiation | 646 (41) | 1,756 (44) | 0.02 |
| Chemotherapy | 53 (3) | 211 (5) | 0.002 |
| Endoscopic pattern | |||
| Remote | 128 (8) | 170 (4) | <0.0001 |
| Recent | 124 (8) | 100 (3) | |
| Sustained | 115 (7) | 66 (2) | |
| Reference | 1,208 (77) | 3,621 (92) | |
| SEER registry region | |||
| Northeast | 339 (22) | 907 (23) | 0.02 |
| Midwest | 231 (15) | 585 (15) | |
| South | 345 (22) | 721 (18) | |
| West | 660 (42) | 1,744 (44) | |
| Education1 | |||
| Q1: 2.5%–13.3% | 388 (25) | 925 (23) | 0.21 |
| Q2: 13.4%–18.4% | 407 (26) | 948 (24) | |
| Q3: 18.5%–22.6% | 395 (25) | 1,047 (26) | |
| Q4: 22.7%–45.7% | 385 (24) | 1,037 (26) | |
| Income2 | |||
| Q1: $18,925–$46,449 | 438 (28) | 937 (24) | 0.005 |
| Q2: $46,450–$52,097 | 352 (22) | 985 (25) | |
| Q3: $52,098–$64,610 | 414 (26) | 1,016 (26) | |
| Q4: $64,611–$91,050 | 371 (24) | 1,019 (26) | |
Data are presented as frequency (%) unless otherwise specified
SD, Standard Deviation
SEER, Surveillance, Epidemiology & End Results
Q, Quartile
Mean % residents in county with college education
Mean county-level median income
Table 2.
Factors associated with localized stage at presentation
| Variable | OR (95%CI) | P value |
|---|---|---|
| Age at diagnosis | 1.00 (0.99, 1.02) | 0.44 |
| Female gender vs. male | 0.90 (0.78, 1.05) | 0.19 |
| Black/Other race vs. white | 0.93 (0.70, 1.24) | 0.64 |
| Charlson comorbidity score | ||
| 0 | Reference | |
| 1 | 1.23 (1.06, 1.42) | 0.01 |
| ≥ 2 | 1.26 (1.00, 1.60) | 0.05 |
| Prior Barrett’s esophagus diagnosis | 2.68 (2.03, 3.56) | <0.0001 |
| Endoscopic pattern | ||
| Reference | Reference | |
| Remote | 1.63 (1.26, 2.11) | <0.0001 |
| Recent | 2.75 (2.06, 3.68) | <0.0001 |
| Sustained | 2.95 (2.07, 4.19) | <0.0001 |
| SEER registry region | ||
| Northeast | Reference | |
| Midwest | 1.08 (0.86, 1.34) | 0.52 |
| South | 1.23 (0.99, 1.53) | 0.06 |
| West | 1.08 (0.91, 1.29) | 0.36 |
| Education 1 | ||
| Q1: 2.5%−13.3% | 0.89 (0.65, 1.21) | 0.26 |
| Q2: 13.4%−18.4% | 1.02 (0.78, 1.35) | 0.40 |
| Q3: 18.5%−22.6% | 0.99 (0.77, 1.28) | 0.78 |
| Q4: 22.7%−45.7% | Reference | |
| Income 2 | ||
| Q1: $18,925-$46,449 | 1.27 (0.92, 1.75) | 0.06 |
| Q2: $46,450-$52,097 | 0.99 (0.74, 1.33) | 0.22 |
| Q3: $52,098-$64,610 | 1.06 (0.82, 1.37) | 0.85 |
| Q4: $64,611-$91,050 | Reference | |
Q, Quartile
Mean % residents in county with college education
Mean county-level median income
The median survival time was 0.56 years in the overall cohort. The median survival was longest in the “sustained” serial endoscopy group at 3.96 years. The reference group median survival was 0.57 years, and that in the “recent” and “remote” groups was 1.81 and 0.89 years, respectively. Kaplan-Meier curves displaying overall survival in the cohort by serial endoscopy group are shown in Figure 4. While better overall survival was observed in all serial endoscopy groups compared with the reference group, the best survival was observed in the “sustained” serial endoscopy group, followed by the “recent” and “remote” groups. The 2-year survival was 62.4% in the “sustained” group, 47.8% in the “recent” group, 31.9% in the “remote” group, and 18.6% in the reference group.
Figure 4.
Kaplan-Meier overall survival curves by serial endoscopic pattern.
A multivariable Cox proportional hazards model was developed to identify factors independently associated with overall survival (Table 3). This model included adjustment for patient demographic factors, SEER registry, and county-level socioeconomic status. A previous diagnosis of Barrett’s esophagus was associated with improved survival (hazard ratio [HR] 0.72, 95% CI 0.62, 0.83). Treatment with radiation or resection was independently associated with improved survival. Compared with the reference group, receipt of serial endoscopy was associated with improved survival. Specifically, consistent with the Kaplan-Meier analysis, the “sustained” serial endoscopy pattern was associated with the greatest impact on survival time compared to the reference group (HR 0.46, 95% CI 0.37, 0.55). The “recent” and “remote” serial endoscopy groups also demonstrated improved survival compared to the reference group, albeit to a lesser degree (HR 0.64, 95% CI 0.55, 0.75 for “recent” and HR 0.87, 95% CI 0.77, 1.00, p=0.04 for “remote”).
Table 3.
Cox proportional hazards model for time to death
| Variable | HR (95%CI) | P value |
|---|---|---|
| Age at diagnosis | 1.03 (1.02, 1.03) | <0.0001 |
| Female gender vs. male | 0.99 (0.92, 1.06) | 0.68 |
| Black/Other race vs. white | 1.02 (0.90, 1.15) | 0.79 |
| Charlson comorbidity score | ||
| 0 | Reference | |
| 1 | 0.93 (0.87, 0.99) | 0.03 |
| ≥ 2 | 1.10 (0.98, 1.22) | 0.10 |
| Prior Barrett’s esophagus diagnosis | 0.72 (0.62, 0.83) | <0.0001 |
| Localized stage vs. other | 0.54 (0.51, 0.58) | <0.0001 |
| Surgery | 0.42 (0.39, 0.46) | <0.0001 |
| Radiation | 0.62 (0.59, 0.66) | <0.0001 |
| Chemotherapy | 0.74 (0.65, 0.84) | <0.0001 |
| Endoscopic pattern | ||
| Reference | Reference | |
| Remote | 0.87 (0.77, 1.00) | 0.04 |
| Recent | 0.64 (0.55, 0.75) | <0.0001 |
| Sustained | 0.46 (0.37, 0.55) | <0.0001 |
| SEER registry region | ||
| Northeast | Reference | |
| Midwest | 1.09 (0.99, 1.20) | 0.09 |
| South | 1.15 (1.04, 1.27) | 0.01 |
| West | 1.03 (0.96, 1.11) | 0.44 |
| Education 1 | ||
| Q1: 2.5%–13.3% | 1.06 (0.92, 1.22) | 0.43 |
| Q2: 13.4%–18.4% | 1.07 (0.94, 1.22) | 0.29 |
| Q3: 18.5%–22.6% | 1.14 (1.01, 1.28) | 0.03 |
| Q4: 22.7%–45.7% | Reference | |
| Income 2 | ||
| Q1: $18,925–$46,449 | 0.99 (0.85, 1.15) | 0.90 |
| Q2: $46,450–$52,097 | 1.03 (0.90, 1.18) | 0.64 |
| Q3: $52,098–$64,610 | 0.94 (0.83, 1.06) | 0.32 |
| Q4: $64,611–$91,050 | Reference | |
Q, Quartile
Mean % residents in county with college education
Mean county-level median income
Lead Time Bias Adjustment
To adjust for lead time bias in screen-detectable cancers, the cohort of 5,532 patients was subdivided into SYM and SCR subgroups based on presence or absence, respectively, of diagnosis codes for symptoms worrisome for malignancy around the time of cancer diagnosis. Among the overall cohort, 25.4% (n=1,406) were in the SCR subgroup. The majority of patients, 74.6% (n=4,126), were in the SYM subgroup. However, the proportion of patients in the SCR subgroup varied by serial endoscopy group, ranging from only 23.2% (n=1,118) in the reference group to 57.5% (n=104) in the sustained group (Table 4). Adjustment for lead time bias was performed (Table 5) assuming a median sojourn time of 2.5 years for patients within the SCR subgroup. Compared with the reference group, receipt of serial endoscopy remained associated with improved survival, with the strongest effect seen in the “sustained” serial endoscopy group (HR 0.45, 95% CI 0.37, 0.55). Sensitivity analyses assuming a median sojourn time of 2 and 3 years had minimal effect on the hazard ratios (Supplementary Table 2).
Table 4.
Proportion of SCR and SYM subgroups within each serial endoscopy group
| Endoscopic pattern | SCR (n=1,406) | SYM (n=4,126) |
|---|---|---|
| Reference | 1,118 (80) | 3,711 (90) |
| Remote | 97 (7) | 201 (5) |
| Recent | 87 (6) | 137 (3) |
| Sustained | 104 (7) | 77 (2) |
Data are presented as frequency (%)
SCR, Screening subgroup
SYM, Symptomatic subgroup
Table 5.
Cox proportional hazards model for time to death with adjustment for lead time bias
| Variable | HR (95%CI) | P value |
|---|---|---|
| Age at diagnosis | 1.03 (1.02, 1.03) | <0.0001 |
| Female gender vs. male | 0.99 (0.92, 1.06) | 0.71 |
| Black/Other race vs. white | 1.02 (0.90, 1.16) | 0.74 |
| Charlson comorbidity score | ||
| 0 | Reference | |
| 1 | 0.93 (0.87, 0.99) | 0.03 |
| ≥ 2 | 1.10 (0.99, 1.22) | 0.09 |
| Prior Barrett’s esophagus diagnosis | 0.74 (0.63, 0.86) | <0.0001 |
| Localized stage vs. other | 0.54 (0.51, 0.58) | <0.0001 |
| Surgery | 0.42 (0.39, 0.46) | <0.0001 |
| Radiation | 0.62 (0.59, 0.66) | <0.0001 |
| Chemotherapy | 0.73 (0.64, 0.83) | <0.0001 |
| Endoscopic pattern | ||
| Reference | Reference | |
| Remote | 0.87 (0.77, 0.99) | 0.03 |
| Recent | 0.63 (0.54, 0.74) | <0.0001 |
| Sustained | 0.45 (0.37, 0.56) | <0.0001 |
| SEER registry region | ||
| Northeast | Reference | |
| Midwest | 1.09 (0.98, 1.20) | 0.10 |
| South | 1.15 (1.04, 1.27) | 0.01 |
| West | 1.04 (0.96, 1.12) | 0.38 |
| Education 1 | ||
| Q1: 2.5%–13.3% | 1.06 (0.92, 1.22) | 0.45 |
| Q2: 13.4%–18.4% | 1.07 (0.94, 1.22) | 0.31 |
| Q3: 18.5%–22.6% | 1.13 (1.01, 1.28) | 0.04 |
| Q4: 22.7%–45.7% | Reference | |
| Income 2 | ||
| Q1: $18,925–$46,449 | 0.99 (0.85, 1.15) | 0.90 |
| Q2: $46,450–$52,097 | 1.04 (0.91, 1.19) | 0.61 |
| Q3: $52,098–$64,610 | 0.94 (0.83, 1.06) | 0.32 |
| Q4: $64,611–$91,050 | Reference | |
Q, Quartile
Mean % residents in county with college education
Mean county-level median income
Length Time Bias Adjustment
To examine length bias, we used 1.5 year survival estimates based on data adjusted for lead time bias (Supplementary Table 3) and examined scenarios where the proportion of aggressive tumors, q, ranged from 0.1 to 0.5, and where θ (the ratio of death rates for less aggressive tumors [B] to more aggressive tumors [A]) ranged from 0.9 to 0.5 (Supplementary Table 4a–d). Conservatively, we used the maximum value of φ across the range of values of θ and q. Because θ values far from 1 are less realistic, we also did this with the same range of q, but for θ = 0.9 to 0.7 and θ = 0.9 to 0.6. We then repeated these calculations under the even more conservative assumption (leading to higher estimates of magnitude of length bias) that no length bias existed in the reference group. Results indicated that there was only a very minimal effect of length bias on the between group relative risk of death (absolute differences of adjusted minus observed risk ratios ≤ 0.01 when using the maximum φ values), for all values of θ (Supplementary Table 5). Under the more conservative assumption of no length bias in the reference group, the differences (length adjusted minus observed risk ratios of death) ranged from 0.04 to 0.06 when θ ranged from 0.5 to 0.9, but were minor (differences of 0.01) for the more realistic scenario where θ ranged from 0.9 to 0.7. Thus the effect of length bias was likely quite small.
Discussion
This population-based study demonstrated that receipt of serial endoscopy prior to esophageal adenocarcinoma diagnosis was associated with improved survival compared with the reference group of patients not receiving serial endoscopy. Among patients receiving serial endoscopy, the “sustained” endoscopy pattern was associated with the greatest improvement in survival.
A prior diagnosis of Barrett’s esophagus was present in 6% (n=329) of the cohort (10% after introduction of a diagnosis code specific for Barrett’s esophagus in 2003). Although the linked SEER-Medicare database does not include pathology data, this finding is consistent with prior studies. A population-based study by Verbeek and colleagues utilizing a nationwide pathology registry in the Netherlands demonstrated that 8% of patients had a diagnosis of Barrett’s esophagus prior to esophageal adenocarcinoma diagnosis [8]. Similarly, a population-based study by Bhat and colleagues from Northern Island revealed that 7% of esophageal adenocarcinoma patients had a prior diagnosis of Barrett’s esophagus [21].
In the current study, patients presenting with localized stage were more likely to have a prior diagnosis of Barrett’s esophagus. Fourteen percent (n=213) of patients with localized stage had a prior BE diagnosis, compared with only 3% (n=116) of patients presenting with other stage. The study by Bhat et al. found that 26% (9/34) of patients with Stage I disease had a prior BE diagnosis, compared with 6% (43/682) of patients presenting with other stage [21]. While patients with a prior diagnosis of BE made up only a small minority of our cohort, taken together, these results provide support for the favorable impact of recognition of at risk patients.
Not surprisingly, patients with localized stage in our study were more likely to undergo surgical resection. However, the overall resection rate (18%) was lower than that (51%) reported in a pooled analysis of 1,148 patients from 4 studies [7], likely reflecting the older population in the current study. Patients with localized stage were more likely to have received serial endoscopy. Receipt of serial endoscopy prior to cancer diagnosis was independently associated with improved survival. Compared with the reference group, the strongest association between serial endoscopy and overall survival was observed in the “sustained” serial endoscopy group (HR 0.46, 95% CI 0.37, 0.55), followed by the “recent” and “remote” serial endoscopy groups. This finding highlights the importance of repeat endoscopic evaluation, particularly over an extended period of time, in patients with risk factors for developing esophageal adenocarcinoma. Based on a prior analysis of the Clinical Outcomes Research Initiative (CORI) database, patients without BE on an initial examination are unlikely to develop BE on a subsequent examination within 5 years [22]. Presumably, those patients in the serial endoscopy groups who did not have BE had other risk factors or symptoms warranting repeat upper endoscopic evaluation. Because we did not have access to their medical records, the indications for these procedures were not available for review. It is also possible that some of the patients in the serial endoscopy groups had BE despite not having an ICD-9-CM diagnosis code for BE. However, as previously mentioned, the proportion of patients in the cohort with BE prior to EAC diagnosis was consistent with that from pathology registry-based studies, suggesting that undercoding of a BE diagnosis was minimal.
Our study had several additional limitations which should be noted. Because we assessed receipt of serial endoscopy in a 5-year period preceding cancer diagnosis using SEER-Medicare data, our cohort was limited to patients aged ≥ 70 years. Given that 68 years is the median age at diagnosis for esophageal cancer in the USA [23], this cohort is somewhat older, limiting the generalizability of our results. To evaluate long-term survival, the last year of diagnosis in the cohort was 2009, so the findings may not reflect current practice. Due to its retrospective nature, the results could be subject to length-time and lead-time bias. Of note, two retrospective studies published within the past several years assessed the impact of a BE diagnosis on esophageal adenocarcinoma survival while adjusting for lead-time and length-time bias. One study demonstrated improved survival among patients with a prior BE diagnosis [21], while the other found that improved esophageal adenocarcinoma survival among patients with a prior BE diagnosis was primarily due to lead-time and length-time bias [11]. As previously noted [24], these disparate results highlight the challenges in adjusting for these biases, since these adjustments are based on certain assumptions which may not be biologically accurate for esophageal adenocarcinoma. Our analysis includes absolute survival rates in each endoscopy group, an outcome which is potentially less susceptible to bias. Moreover, adjustment for lead time and length time bias did not result in any substantive change in our findings.
In conclusion, receipt of more than one endoscopy prior to esophageal adenocarcinoma diagnosis was associated with improved survival, particularly in patients with a “sustained” endoscopy pattern. Improved survival persisted even after adjustment for lead time and length time bias. Consistent with prior studies, patients with a prior diagnosis of Barrett’s esophagus made up a minority of the cohort, but were more likely to present with localized stage. These findings highlight the importance of early recognition of high risk individuals.
Supplementary Material
Acknowledgments
This study used the linked SEER-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the Applied Research Program, NCI; the Office of Research, Development and Information, CMS; Information Management Services (IMS), Inc.; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database.
Grant support:
This study was funded by an American College of Gastroenterology Junior Faculty Development Grant awarded to L.C. Cummings. A.C. was supported by NIH grants U54CA163060 (Case Barrett’s Esophagus Translational Research Network) and P50 CA150964 (Case GI SPORE). G.S.C. was supported by NIH grant P50 CA150964 (Case GI SPORE), UL1 TR000439 (Case Clinical & Translational Science Collaborative), and P30 CA043703 (Case Comprehensive Cancer Center). M.S. was supported by P30 CA043703 (Case Comprehensive Cancer Center).
Footnotes
Financial Disclosures: None
References
- 1.Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer Statistics, 2021. CA Cancer J Clin. 2021;71(1):7–33. [DOI] [PubMed] [Google Scholar]
- 2.Rustgi AK, El-Serag HB. Esophageal Carcinoma. N Engl J Med. 2014;371(26):2499–509. [DOI] [PubMed] [Google Scholar]
- 3.Lagergren J, Lagergren P. Recent Developments in Esophageal Adenocarcinoma. CA Cancer J Clin. 2013;63:232–48. [DOI] [PubMed] [Google Scholar]
- 4.AGA Institute Medical Position Panel. American Gastroenterological Association medical position statement on the management of Barrett’s esophagus. Gastroenterology. 2011;140:1084–91. [DOI] [PubMed] [Google Scholar]
- 5.Shaheen NJ, Falk GW, Iyer PG, Gerson LB. ACG Clinical Guideline: Diagnosis and Management of Barrett’s Esophagus. Am J Gastroenterol. 2016;111(1):30–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Standards of Practice Committee of the American Society for Gastrointestinal Endoscopy. ASGE guideline on screening and surveillance of Barrett’s esophagus. Gastrointest Endosc. 2019;90(3):335–59. [DOI] [PubMed] [Google Scholar]
- 7.Codipilly DC, Chandar AK, Singh S, et al. The Effect of Endoscopic Surveillance in Patients With Barrett’s Esophagus: A Systematic Review and Meta-analysis. Gastroenterology. 2018;154(8):2068–86. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Verbeek RE, Leenders M, Ten Kate FJW, et al. Surveillance of Barrett’s Esophagus and Mortality from Esophageal Adenocarcinoma: A Population-Based Cohort Study. Am J Gastroenterol. 2014;109(8):1215–22. [DOI] [PubMed] [Google Scholar]
- 9.El-Serag HB, Naik AD, Duan Z, et al. Surveillance endoscopy is associated with improved outcomes of oesophageal adenocarcinoma detected in patients with Barrett’s oesophagus. Gut. 2016;65(8):1252–60. [DOI] [PubMed] [Google Scholar]
- 10.Corley DA, Mehtani K, Quesenberry C, Zhao W, de Boer J, Weiss NS. Impact of Endoscopic Surveillance on Mortality From Barrett’s Esophagus–Associated Esophageal Adenocarcinomas. Gastroenterology. 2013;145(2):312–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Tramontano AC, Sheehan DF, Yeh JM, et al. The Impact of a Prior Diagnosis of Barrett’s Esophagus on Esophageal Adenocarcinoma Survival. Am J Gastroenterol. 2017;112:1256–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Old O, Moayyedi P, Love S, et al. Barrett’s Oesophagus Surveillance versus endoscopy at need Study (BOSS): protocol and analysis plan for a multicentre randomized controlled trial. J Med Screen. 2015;22(3):158–64. [DOI] [PubMed] [Google Scholar]
- 13.Cooper GS, Kou TD, Chak A. Receipt of previous diagnoses and endoscopy and outcome from esophageal adenocarcinoma: a population-based study with temporal trends. Am J Gastroenterol. 2009;104:1356–62. [DOI] [PubMed] [Google Scholar]
- 14.Surveillance, Epidemiology, and End Results Database. Available at: http://seer.cancer.gov/about/factsheets/SEER_Overview.pdf. Accessed October 9, 2019.
- 15.Collaborative Stage for TNM 7, Esophagus Histologies Stage Table. Available at: http://web2.facs.org/cstage0204/esophagus/Esophagus_xhw.html. Accessed October 28, 2013.
- 16.Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45:613–9. [DOI] [PubMed] [Google Scholar]
- 17.NCI Comorbidity Index Overview. Available at: https://healthcaredelivery.cancer.gov/seermedicare/considerations/comorbidity.html. Accessed February 3, 2021.
- 18.Cook MB, Drahos J, Wood S, et al. Pathogenesis and progression of oesophageal adenocarcinoma varies by prior diagnosis of Barrett’s oesophagus. Br J Cancer. 2016;115:1383–90. https://www.nature.com/articles/bjc2016344#supplementary-information. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Cummings LC, Kou TD, Schluchter MD, Chak A, Cooper GS. Outcomes after endoscopic versus surgical therapy for early esophageal cancers in an older population. Gastrointest Endosc. 2016;94(2):232–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Duffy SW, Nagtegaal ID, Wallis M, et al. Correcting for Lead Time and Length Bias in Estimating the Effect of Screen Detection on Cancer Survival. Am J Epidemiol. 2008;168(1):98–104. [DOI] [PubMed] [Google Scholar]
- 21.Bhat SK, McManus DT, Coleman HG, et al. Oesophageal adenocarcinoma and prior diagnosis of Barrett’s oesophagus: a population-based study. Gut. 2015;64(1):20–5. [DOI] [PubMed] [Google Scholar]
- 22.Rodriguez S, Mattek N, Lieberman D, Fennerty B, Eisen G. Barrett’s Esophagus on Repeat Endoscopy: Should We Look More Than Once? Am J Gastroenterol. 2008;103:1892–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.SEER Cancer Statistics Factsheets: Esophageal Cancer. Available at: http://seer.cancer.gov/statfacts/html/esoph.html. Accessed May 14, 2018.
- 24.Zakko L, Visrodia K, Wang KK, Iyer PG. Editorial: The Effect of Bias on Estimation of Improved Survival After Diagnosis of Barrett’s Esophagus. Am J Gastroenterol. 2017;112:1265–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.