Association of the interval between endoscopies with gastric cancer stage at diagnosis in a region of high prevalence

Authors


Abstract

BACKGROUND:

Endoscopy screening in high-risk populations may reduce gastric cancer mortality by detecting cancer earlier. We evaluated the association between the interval between upper gastrointestinal endoscopies and the gastric cancer stage at diagnosis in patients from a region of high prevalence.

METHODS:

The study cohort consisted of 2485 patients diagnosed with gastric adenocarcinoma. We evaluated the effect on cancer stage of the interval between the endoscopy that was diagnostic for gastric cancer and the endoscopy preceding it. Patients were stratified into 7 groups: 1-, 2-, 3-, 4-, 5-, >5-year intervals and those who were never screened.

RESULTS:

The risk of higher cancer stage at diagnosis increased by 23% per increase in interval length (odds ratio = 1.23, 95% confidence interval [CI] = 1.19-1.28). Compared to the never-screened, the odds ratio of having a higher stage of cancer decreased gradually from 0.53 (95% CI = 0.41-0.69) in the >5-year interval group to 0.31 (95% CI = 0.24-0.40) in the 1-year interval group. Compared to the 1-year interval group, the risk of advanced gastric cancer was increased in the 4- and 5-year, but not the 2- and 3-year, interval groups. However, patients with a family history of gastric cancer were more likely to have a higher stage at diagnosis if they had a 3-year interval rather than a 1-year interval.

CONCLUSIONS:

A significant benefit in cancer stage at diagnosis was observed in all interval groups compared to never-screened. Endoscopy intervals of 3 years or less showed similar benefits, but family members of gastric cancer patients may benefit from intervals of under 3 years. Cancer 2012. © 2012 American Cancer Society.

INTRODUCTION

Although the incidence of gastric cancer has decreased substantially in the Western world, it is still the fourth most common cancer and the second leading cause of mortality among cancers.1 The highest incidence rates are in Eastern Asia, including Korea, Japan, and China.1 The prognosis of gastric cancer strongly depends on the stage of the cancer at the time of diagnosis.2 The high mortality from gastric cancer is mainly due to the late presentation of the disease. Early detection and treatment is the only way to reduce mortality.3 The early detection of gastric cancer allows for curative therapy and better quality of life after treatment. Further, early detection allows less-invasive treatments, such as endoscopic submucosal dissection and laparoscopic gastrectomy.4, 5

Screening methods for gastric cancer include photofluorography, upper gastrointestinal series, serum pepsinogen testing, Helicobacter pylori antibody testing, and upper gastrointestinal endoscopy.3, 6, 7 In Japanese studies, the risk of death from gastric cancer was significantly lower in patients screened with photofluorography than in unscreened patients.8, 9 Recently, we reported that endoscopic screening repeated every 2 years decreased the incidence of gastric cancer and the number of cases diagnosed at advanced stages.10 There have been inconsistent findings on the effectiveness of endoscopy screening in reducing gastric cancer–related mortality, most likely due to variability in the screening intervals of the studies.11, 12 The aim of this study was to evaluate whether there is an association between the interval between upper endoscopies and the stage of gastric cancer at the time of diagnosis, which is the most powerful predictor of patient survival.

MATERIALS AND METHODS

Study Design and Population

Consecutive patients who presented to the Center for Gastric Cancer at the Korean National Cancer Center in Goyang, Korea, between January 2004 and December 2009 were eligible for the study. All participants were examined by upper endoscopy and had the diagnosis of gastric adenocarcinoma confirmed pathologically. Information relating to the patients' last endoscopy before gastric cancer diagnosis and their baseline characteristics was obtained via prospectively collected questionnaires. Exclusion criteria were as follows: those who had been already diagnosed with or suspected to have a gastric cancer or adenoma, those who underwent endoscopy <6 months before the diagnosis of cancer, those who had a history of gastric surgery before the diagnosis of cancer, and those who were found to have multiple synchronous gastric cancers (because the individual cancers could have been at a different tumor stage). This retrospective study was approved by the Institutional Review Board of the National Cancer Center (NCCNCS-11-467).

Endoscopy Interval and Cancer Stage

We divided the patients into the following 7 groups on the basis of the interval between the endoscopy that led to the diagnosis of gastric cancer and the endoscopy that preceded it (interval between endoscopies): 1-year (7-18 months), 2-year (19-30 months), 3-year (31-42 months), 4-year (43-54 months), 5-year (55-66 months), and >5-year (>66 months) interval groups, and those with no previous endoscopy (never-screened). The stage of gastric cancer was defined by American Joint Committee on Cancer (AJCC) 6th edition13 and also classified as early (EGC) or advanced gastric cancer (AGC). EGC was defined as carcinoma confined to the mucosa or submucosa regardless of lymph node metastasis (T1 in T classification), and AGC was defined as carcinoma that involved the proper muscle layer or beyond (T2 or greater). For the grouping according to tumor location, the Japanese classification of gastric carcinoma, 13th edition, was used, with “proximal” indicating that the tumor was located in the upper one-third of the stomach and “distal” indicating that the tumor was in the lower two-thirds of the stomach.14

Statistical Analysis

One-way analysis of variance was used for the comparison of the mean ages, and Pearson's chi-square test was used to compare differences for categorical variables. The Spearman correlation test was used to test the correlation between cancer stage and the interval between endoscopies (months) in patients with a previous history of endoscopy. To analyze the relationships between the 7 endoscopy interval groups and the stages of gastric cancer, ordinal logistic regression was used, and odds ratios (ORs) and 95% confidence intervals (95% CIs) were determined. The endoscopy interval was the independent variable and was adjusted for age and sex. We also estimated the OR for each endoscopy interval group compared to the never-screened group by logistic regression after adjusting for age and sex. Logistic regression and P for trend were used to estimate the ORs for AGC and EGC compared to the 1-year interval group. Logistic regression analysis was performed without the use of an elimination method. We stratified analyses by age, sex, family history of gastric cancer (defined as cancer in first- or second-degree relatives), tumor location, histological tumor differentiation, and Lauren's classification15 to investigate which groups would benefit from frequent endoscopy. In the subgroup analysis, ordinal logistic regression was used to analyze the relationships between the 7 endoscopy interval groups and gastric cancer stages. All P values are 2-sided, and significance is indicated by P < .05. All statistical analyses were performed using STATA software, version 10.1 (StataCorp, College Station, Tex).

RESULTS

Patients and Baseline Demographic Characteristics

Of the screened 2668 patients with gastric cancer, 183 subjects were excluded from the study because of the following reasons: a previous diagnosis of gastric adenocarcinoma (n = 6) or gastric adenoma (n = 19); a previous history of gastric resection (n = 7); an endoscopy was performed on the patient within 6 months of diagnosis (n = 6); the lack of pathological confirmation of gastric adenocarcinoma (n = 13); or the presence of multiple gastric cancers (n = 132).

The baseline demographics and tumor characteristics of the 2485 enrolled patients are shown in Table 1. The mean age was lower in patients with stage IV cancer (55.4 ± 12.1 years) than in those with either stage I (58.0 ± 11.5 years; P < .001) or stage II (58.1 ± 11.4 years; P = .029) cancer. The male-to-female ratio was 2 to 1. There were no differences in the stage of cancer at the time of diagnosis between sexes (P = .481). In addition, there were no significant differences among the stages of cancer in terms of smoking and drinking status, and education level (P = .263, P = .182, and P = .128, respectively). Helicobacter pylori infection rate was 78.8%, and it did not differ among the stages of cancer (P = .848).

Table 1. Baseline Demographics and Tumor Characteristics
Characteristicn = 2485%
  • a

    Undetermined 153 patients were missed.

  • b

    First- or second-degree family history (553 and 123 patients, respectively).

  • c

    Cancer stage (I-IV) according to American Joint Committee on Cancer 6th edition.

  • d

    Undetermined 168 patients were missed.

  • e

    Includes 130 patients with mixed type.

Age at diagnosis, y (mean age: 57.3±11.9 y, range: 21-84 y)
 <401998.0
 40-4949219.8
 50-5961724.8
 60-6977931.4
 ≥7039816.0
Sex  
 Male165766.7
 Female82833.3
Smoking  
 No99640.1
 Yes148959.9
Drinking  
 No84734.1
 Yes163865.9
Education (year)  
 <12127351.2
 ≥12121248.8
Helicobacter pylori infectiona  
 No37315.0
 Yes195978.8
Family history of gastric cancerb  
 No180972.8
 Yes67627.2
Tumor stagec  
 I139956.3
 II27411.0
 III2439.8
 IV56922.9
Tumor location  
 Proximal40916.5
 Distal193477.8
 Whole stomach1425.7
Histological differentiation  
 Differentiated101740.9
 Undifferentiated146859.1
Lauren classificationd  
 Intestinal120748.6
 Diffuse and mixede111044.7

Association Between Interval Between Endoscopies and Cancer Stage

The median interval between endoscopies was 30 months (interquartile range: 16.5-62.0 months) in the 1216 patients who had undergone previous upper endoscopy at least once. An increased interval between endoscopies was significantly associated with higher gastric cancer stage (P < .001). After adjusting for age and sex, the risk of having a higher stage of cancer increased by 1.0023 times (95% CI = 1.0004-1.0042, P = .013) and the likelihood of detecting AGC increased by 1.003 times (95% CI = 1.001-1.005, P = .002) for every increased month of endoscopy interval.

The percentage of patients with stage I gastric cancer was similar (approximately 70%) in the 3 groups with intervals between endoscopies of 3 years or less, whereas it was 45.5% in the never-screened group (Fig. 1). Overall, a shift toward a higher stage at diagnosis was observed in patients with longer intervals between endoscopies. For every increase in interval length (and thus, for each interval group from 1-year to never-screened), the risk of having a higher stage at diagnosis increased by 23% after adjusting for age and sex (OR = 1.23, 95% CI = 1.19-1.28, P < .001).

Figure 1.

Distribution of gastric cancer stages is shown according to the interval between endoscopies. EGD indicates esophagogastroduodenoscopy.

Next, we analyzed whether the 6 screened groups had lower cancer stages at diagnosis than the never-screened group (Table 2). A significant advantage was observed in all 6 interval groups. The OR for higher stage cancer at diagnosis was 0.53 in the >5-year interval group, and the OR gradually decreased as the interval to endoscopy decreased, reaching 0.31 in the 1-year interval group.

Table 2. The Odds Ratios from an Ordinal Logistic Model With Increasing Tumor Stagea as the Dependent Variable
Previous EndoscopyInterval (y)bNo. of CasesORc95% CIP
  • CI indicates confidence interval; NA, not applicable; OR, odds ratio.

  • a

    Cancer stage (I-IV) according to American Joint Committee on Cancer 6th edition.

  • b

    Time interval between the endoscopy that led to the diagnosis of gastric cancer and the preceding endoscopy.

  • c

    ORs for ordinal logistic regression were adjusted for age and sex.

NoNA12681.0 Reference
Yes>52800.530.41-0.69<.001
 5880.510.33-0.78.002
 4910.440.29-0.68<.001
 31540.360.25-0.51<.001
 22680.330.25-0.44<.001
 13360.310.24-0.40<.001

Optimal Endoscopy Interval

We evaluated the risk of detecting AGC rather than EGC in each interval group (Table 3). Within the 1- to 5-year interval groups, the ORs for the detection of AGC tended to increase with an increasing interval, yielding a P for trend of .001. The risk of AGC was not increased in patients in the 2- or 3-year interval group (OR = 1.11, and OR = 1.21, respectively; both P > .05), whereas it was increased in the 4- or 5-year interval groups.

Table 3. Odds Ratios for the Detection of Advanced Gastric Cancera in the Different Interval Groups Compared to the 1-Year Interval Group
Interval (y)bNo. of CasesORc95% CIPP for Trend
  • CI indicates confidence interval; OR, odds ratio.

  • a

    Depth of tumor invasion involved the proper muscle layer or beyond.

  • b

    Time interval between the endoscopy that led to the diagnosis of gastric cancer and the patient's preceding endoscopy.

  • c

    ORs for logistic regression were adjusted for age and sex.

13361.0 Reference 
22681.110.75-1.64.600 
31541.210.75-1.97.432.001
4912.531.41-4.55.001 
5882.161.26-3.69.004 

Analyses According to Demographic Characteristics and Tumor Subtypes

The subgroup analysis based on the baseline demographics and tumor characteristics of the patients is presented in Figure 2A. The ORs of having a higher stage of cancer at diagnosis were significantly increased by having a longer interval between endoscopies in all age subgroups (P < .001) except the <40-years group (OR = 1.08, 95% CI = 0.96-1.23, P = .192). The ORs were also significant in most subgroups irrespective of sex, family history of gastric cancer, tumor location and differentiation, or Lauren classification. The subgroup of whole stomach was excluded from the analysis because of a small sample size.

Figure 2.

Odds ratios (ORs) obtained from multivariate ordinal logistic regression analysis of the relationship between the interval between endoscopies and tumor stage are shown for (A) all 2485 patients and (B) the 758 patients who underwent a previous endoscopy within a 3-year interval. CI indicates confidence interval.

Next, we evaluated whether there is a subgroup for which an interval between endoscopies of <3 years is beneficial. In the analysis of 758 patients who underwent the last previous endoscopy within 3 years of diagnosis, the ORs for a higher stage of cancer were not significantly different between interval groups in most subgroups (Fig. 2B). However, patients with a family history of gastric cancer did show a benefit from more frequent endoscopies (OR = 1.62, 95% CI = 1.10-2.37, P = .014), whereas patients without a family history of gastric cancer did not show benefit (OR = 0.95, 95% CI = 0.74-1.21, P = .657). In addition, the ORs for having a higher cancer stage was significantly associated with a longer interval between endoscopies in patients aged 60 to 69 years and patients with intestinal type cancer (OR = 1.45, P = .041 and OR = 1.52, P = .010, respectively).

In family members of patients with gastric cancer, an interval of endoscopy of 3 years significantly increased the risk of having a higher tumor stage, compared to the 1-year interval group (OR = 2.68, 95% CI = 1.26-5.67, P = .010). This increase in risk was not observed in the 2-year interval group with a family history of gastric cancer (OR = 1.88, 95% CI = 0.97-3.66, P = .061) and the 2- and 3-year interval groups without a family history. In patients with family history, the risk of having a higher stage showed marginal significance in the 3-year interval group compared with the 2-year interval group (OR = 2.05, 95% CI = 0.97-4.32, P = .059). In patients in their 60s, the risk of having a higher stage was increased in the 3-year interval group, but not the 2-year interval group, when compared to the 1-year interval group (OR = 2.09, 95% CI = 1.00-4.34, P = .049). In patients with intestinal type cancer, both the 2- and 3-year interval group showed a significantly increased risk of having a higher tumor stage (OR = 1.89, 95% CI = 1.08-3.30, P = .027 and OR = 2.22, 95% CI = 1.15-4.32, P = .018, respectively).

DISCUSSION

In this study, we confirmed a significant benefit in all interval groups compared to never-screened patients in cancer stage at diagnosis. We determined that endoscopy intervals of 3 years or less were associated with an earlier stage of gastric cancer at diagnosis. However, patients with a family history of gastric cancer and patients in their 60s showed a benefit from more frequent endoscopies.

Mass screening for gastric cancer with photofluorography was initiated in Japan in 1960, and patients diagnosed with gastric cancer via screening were more likely to have an earlier stage of cancer at diagnosis and a better survival rate than those diagnosed by symptoms.16 Gastric cancer is the most common cancer in Korea, and biennial gastric cancer screening via either upper gastrointestinal series or upper endoscopy has been conducted in Korea nationwide since 1999 for individuals aged 40 years and older.17 A few Asian studies have reported that screening endoscopies repeated every 2 years increased the early detection rate of gastric cancer.10, 18 In contrast, no reduction in gastric cancer mortality was observed in individuals screened every 5 years.12 Thus, the optimal frequency of endoscopy for gastric cancer screening remains uncertain. A randomized, controlled trial that measures the effects of gastric cancer screening on mortality may be difficult and unethical, even in a region of high prevalence. However, considering that stage of gastric cancer is well correlated with the survival rate, our study suggests that early detection of gastric cancer by screening may increase patient survival.

A recent population-based cohort study that used photofluorography screening showed that although the overall incidence rate did not differ between the screened and unscreened individuals (relative risk = 1.06, 95% CI = 0.90-1.25), there was a significant decrease in the incidence of AGC in screened individuals (relative risk = 0.75, 95% CI = 0.58-0.96).9 Another study conducted on patients with cardia adenocarcinoma found that 49% of the cancer patients who underwent previous endoscopy presented with in situ or local stage cancer, compared with 27% of those who did not have a previous endoscopy.19 These findings are compatible with our study and are consistent with the notion that endoscopic screening, especially when done frequently, helps lead to earlier diagnosis of gastric cancer.

We evaluated who might benefit from frequent endoscopy by stratifying the patients according to age, sex, family history of gastric cancer, and cancer subtypes. A shift toward a higher cancer stage at diagnosis in patients with longer interval between endoscopies was observed in all patients irrespective of baseline demographics and cancer subtypes, except in patients younger than 40 years. This discrepancy according to age is probably due to the different clinical characteristics of gastric cancers in younger patients, specifically, a more aggressive clinical course and a tendency to develop a more advanced stage of cancer.20 Our results support this, because we found that among patients under 40 years of age, 32% and 30% of patients in the 1- and 2-year interval groups, respectively, were confirmed to have stage IV gastric cancer at diagnosis. The finding that the patients with stage IV disease were younger than the patients with earlier stage disease leads to the perception that younger patients have more rapidly progressive disease and thus should likely be screened more often. However, it is possible that these advanced stage cancers might have been interval cancers that grew rapidly; thus, even if more screenings were conducted, the disease would not have been detected any earlier. Further, because the incidence of gastric cancer is very low in young individuals, screening this population more frequently, even in regions with high risk for gastric cancer, does not seem to be cost-effective. In addition, some young patients with the linitis plastica type of gastric cancer might have undergone endoscopy for cancer symptoms 1 to 2 years earlier, when the gastric cancer was missed. Because the gastric mucosa is frequently spared from malignant invasion in linitis plastica, endoscopic diagnosis can be very difficult.21 A retrospective study reported that the typical features of linitis plastica were found in only 11.8% of upper endoscopies for patients with linitis plastica.22

A positive family history of gastric cancer has been identified as an important risk factor for gastric cancer,23 and gastric cancer guidelines refer to individuals with a family history of gastric cancer as a high-risk population.24, 25 However, it remains unclear to what extent screening programs are effective, especially with regard to whether more frequent screening is appropriate for family members of gastric cancer patients. We found that in patients with a family history of gastric cancer, the risk of a higher stage of gastric cancer at diagnosis was greater in patients who had a 3-year interval between endoscopies than in those with a 1-year interval and probably than in those with a 2-year interval, whereas it was not increased in those without a family history of gastric cancer. We also previously showed that tumors in patients with an affected first-degree relative tended to be larger or more deeply infiltrated than the tumors of patients without a family history, even when the patients had undergone an endoscopy within 2 years of diagnosis.10 Thus, our data suggest that family members of gastric cancer patients should be considered for more frequent endoscopies than those without a family history, in order to detect gastric cancer at an earlier stage.

In addition, shorter intervals between endoscopies (<3 years) also showed a benefit in terms of stage at diagnosis in patients in their 60s and in patients with intestinal-type gastric cancer. The benefit in these subgroups may be associated with the so-called length bias, in which a screening test can detect less aggressive tumors with a better prognosis more frequently than the more aggressive tumors.26 The clinical course of gastric cancer is less aggressive in older patients than in younger patients, and intestinal-type gastric cancer tends to grow more slowly and have a better prognosis than diffuse-type cancer.27, 28 However, it is possible that the finding that only patients in their 60s benefited from an interval between endoscopies of <3 years is a spurious finding. Further studies are needed to clarify this point, because the ORs gradually increased as age increased.

There are several limitations to this study. First, this is a single-center study in a tertiary cancer center, and thus we cannot exclude the possibility of selection bias. Second, because the information relating to the previous endoscopy was self-reported, these data were vulnerable to recall bias. Our findings should be confirmed by future studies based on objectively evaluated endoscopy interval data. Third, because the medical records of the results of previous endoscopies were not obtained, all data related to the histological changes, including glandular atrophy and intestinal metaplasia, and the H. pylori infection status at the time of any previous endoscopy were not available. Fourth, gastric cancer screening efficacy cannot be directly estimated by cancer stage, because there might be a gap related to lead-time bias and length bias between cancer stage and mortality.26 However, gastric cancer stage is well associated with survival rate.29, 30 Moreover, although EGC has a relatively long natural history, it progresses to an advanced stage and leads to death from gastric cancer in most untreated patients.31 Fifth, we did not investigate gastric mucosal status. Hence, there is no knowledge of factors such as glandular atrophy and intestinal metaplasia, both of which can increase gastric cancer risk and whose presence leads to recommendation of shorter intervals between endoscopies.32 Sixth, it is likely that some participants underwent the previous upper endoscopy due to symptoms rather than for the purpose of screening. However, the exclusion of prediagnosis endoscopies within 6 months could minimize the possibility of including individuals who needed clinical assessment due to gastric cancer. Seventh, socioeconomic status, diet, smoking, and other lifestyle factors may influence not only the development of gastric cancer but also potentially the speed with which tumors develop. Some patients who have had annual endoscopy screening due to employer-sponsored programs may have health habits that predispose them to better outcome. In addition, those who were screened less frequently than the recommended 2-year interval were probably more likely to have other characteristics that predisposed to poorer health outcomes in general. These are potentially significant confounders, and are a limitation to this study. Finally, our data are from an area in which there is a high prevalence of gastric cancer. As such, this study might not apply to the Western world, where the gastric cancer risk is much lower.

This study has the following implications. First, intervals between prediagnosis upper gastrointestinal endoscopies of 3 years or less were associated with an earlier stage of gastric cancer at diagnosis in all cancer patients except those younger than 40 years. Thus, our results indicate that individuals older than 40 years with average risk factors and who live in a high-risk region for gastric cancer should be considered for screening endoscopy at least every 3 years in order to detect gastric cancer at an earlier stage. Also, our data suggest that changing the current recommendation of biennial gastric cancer screening in Korea to one of screening every 3 years should be considered. Second, high-risk individuals, such as those with a family history of gastric cancer and/or who are aged 60 years or older, might be considered for endoscopy screening with intervals of under 3 years. Further well-designed studies are needed to confirm the findings from this observational study.

FUNDING SOURCES

This work was supported by a grant from the National Cancer Center, Korea (NCC 1010190).

CONFLICT OF INTEREST DISCLOSURE

The authors made no disclosure.

Ancillary