ORIGINAL ARTICLE

Endoscopist biopsy rate as a quality indicator for outpatient

gastroscopy: a multicenter cohort study with validation
Wladyslaw Januszewicz, MD,1,2,3 Paulina Wieszczy,2,4 Andrzej Bialek, MD,5 Katarzyna Karpinska, MD,6
Jakub Szlak, MD,1 Jakub Szymonik, MD,1 Maciej Rupinski, MD,1,2 Andrzej Mroz, MD, PhD,7,8
Jaroslaw Regula, MD, PhD,1,2 Michal F. Kaminski, MD, PhD1,2,4,9
Warsaw, Poland; Cambridge, United Kingdom; Szczecin, Poland; Oslo, Norway

Background and Aims: The diagnosis of gastric premalignant conditions (GPCs) relies on endoscopy with
mucosal sampling. We hypothesized that the endoscopist biopsy rate (EBR) might constitute a quality indicator
for EGD, and we have analyzed its association with GPC detection and the rate of missed gastric cancers (GCs).
Methods: We analyzed EGD databases from 2 high-volume outpatient units. EBR values, defined as the proportion
of EGDs with
1 biopsy to all examinations were calculated for each endoscopist in Unit A (derivation cohort) and
divided by the quartile values into 4 groups. Detection of GPC was calculated for each group and compared using
multivariate clustered logistic regression models. Unit B database was used for validation. All patients were followed
in the Cancer Registry for missed GCs diagnosed between 1 month and 3 years after EGDs with negative results.
Results: Sixteen endoscopists in Unit A performed 17,490 EGDs of which 15,340 (87.7%) were analyzed. EBR
quartile values were 22.4% to 36.7% (low EBR), 36.8% to 43.7% (moderate), 43.8% to 51.6% (high), and 51.7%
and 65.8% (very-high); median value 43.8%. The odds ratios for the moderate, high, and very-high EBR groups
of detecting GPC were 1.6 (95% confidence interval [CI], 1.3-1.9), 2.0 (95% CI, 1.7-2.4), and 2.5 (95% CI, 2.1-
2.9), respectively, compared with the low EBR group (P < .001). This association was confirmed with the
same thresholds in the validation cohort. Endoscopists with higher EBR (
43.8%) had a lower risk of missed can-
cer compared with those in the lower EBR group (odds ratio, 0.44; 95% CI, 0.20-1.00; P Z .049).
Conclusions: The EBR parameter is highly variable among endoscopists and is associated with efficacy in GPC
detection and the rate of missed GCs. (Gastrointest Endosc 2019;-:1-9.)

INTRODUCTION incidence and mortality of GC, it remains the fifth most

common malignancy in the world and the third leading
Gastric cancer (GC) usually presents at an advanced cause of cancer mortality.1 In 2015, there were 1.3
stage in the Western world with limited opportunities for million incident cases and 819,000 deaths due to GC
curative therapy. Despite a consistent decline in the global worldwide.2

Abbreviations: BE, Barrett’s esophagus; CI, confidence interval; DA, Education, Warsaw, Poland (2); MRC Cancer Unit, University of
duodenal adenoma; EBR, endoscopist biopsy rate; GC, gastric cancer; Cambridge, Cambridge, United Kingdom (3); Department of Cancer
GD, gastric dysplasia; GPC, gastric precancerous condition; OR, odds ra- Prevention, The Maria Sklodowska-Curie Memorial Cancer Center and
tio; SIN, squamous intraepithelial neoplasm; UGI, upper Institute of Oncology, Warsaw, Poland (4); Department of
gastrointestinal. Gastroenterology, Pomeranian Medical University, Szczecin, Poland (5);
Department of Pathomorphology, Pomeranian Medical University,
DISCLOSURE: All authors disclosed no financial relationships relevant
Szczecin, Poland (6); Department of Pathomorphology, Medical Centre
to this publication.
for Postgraduate Education, Warsaw, Poland (7); Department of
Copyright ª 2019 by the American Society for Gastrointestinal Endoscopy Pathology and Laboratory Medicine, The Maria Sklodowska-Curie
0016-5107/$36.00 Memorial Cancer Centre and Institute of Oncology, Warsaw, Poland (8);
https://doi.org/10.1016/j.gie.2019.01.008 Department of Health Management and Health Economics, University of
Oslo, Oslo, Norway (9).
Received September 10, 2018. Accepted January 2, 2019.
Reprint requests: Wladyslaw Januszewicz, MD, MRC Cancer Unit, University
Current affiliations: Department of Gastroenterological Oncology, The
of Cambridge, Cambridge CB2 0XZ, United Kingdom.
Maria Sklodowska-Curie Memorial Cancer Centre and Institute of
Oncology, Warsaw, Poland (1); Department of Gastroenterology, If you would like to chat with an author of this article, you may contact
Hepatology and Clinical Oncology, Medical Centre for Postgraduate Dr Januszewicz at wj264@mrc-cu.cam.ac.uk.

www.giejournal.org Volume -, No. - : 2019 GASTROINTESTINAL ENDOSCOPY 1

Biopsy rate as a gastroscopy quality marker Januszewicz et al

A significant proportion of these neoplasms arise from viewed and accepted by the ethics committee at the
benign, precancerous conditions. For instance, the most authors’ institution on June 14, 2017 (49/PB/2017).
common subtype of GC, a non-cardia intestinal-type adeno-
carcinoma, develops on a background of long-standing Endoscopist biopsy rate parameter
mucosal inflammation through a number of stages from For each endoscopist, we counted biopsy rate (EBR) as
chronic atrophic gastritis, by way of intestinal metaplasia, a percentage of EGDs with at least 1 biopsy specimen ob-
through low-grade and high-grade dysplasia, up to cancer. tained for histology from the esophagus, stomach, or the
This sequence is known as the Correa cascade.3 duodenum (ICD-9 codes 45.14, 45.16, 42.24) for all
EGD with biopsies has a primary role in the diagnosis and EGDs performed in the study period. Biopsies for rapid
surveillance of patients with gastric precancerous conditions urease test were not included.
(GPCs) and is considered to have high sensitivity and spec-
ificity in cancer diagnosis. However, despite increasing Derivation group (Unit A)
experience in the field of endoscopy, a significant propor- In the analysis, we have included full reports on consec-
tion of neoplastic lesions remain undetected. A recent utive adult patients undergoing diagnostic EGD, mostly for
meta-analysis has shown that 11.3% of upper gastrointestinal evaluation for UGI symptoms. All endoscopists included in
(UGI) tract cancers are missed at EGD up to 3 years before the study were staff specialists in internal medicine or
the diagnosis.4 These findings underscore the importance gastroenterology, had undergone dedicated training in
of quality control for this procedure. UGI endoscopy, and were considered competent to
In recent years, several guidelines and position state- perform diagnostic EGDs independently. Reports on juve-
ment papers on EGD quality have been published.5-7 nile patients (<18 years old) or those with incomplete data
Most of the presented measures, however, are based on were excluded; EGDs performed by trainees and endo-
low-quality evidence and are not associated with significant scopists who did fewer than 100 procedures during the
outcomes, such as neoplasia detection or interval cancer study period were also excluded. Unit A database consisted
risk. Because the diagnosis of precancerous conditions of endoscopy reports with
and early cancers in conventional EGD relies on biopsy  Patient data: personal identification number (PESEL),
sampling of suspicious areas in the UGI tract, we hypothe- hospital number (PID), gender, and date of birth;
sized that within a group of competent endoscopists, the  EGD data: indication, date, type of sedation, type
rate of obtaining biopsy specimens during endoscopy (model) of endoscope, examination report (descriptive),
broadly reflects the number of detected abnormalities. main gastroscopy findings, procedure ICD-9 code, num-
We considered this as a potentially objective and reproduc- ber and description of containers used for histopatholo-
ible quality marker for routine, outpatient EGD. Therefore, gy specimens, and rapid urease test (if obtained);
the aim of this study was to investigate the variability in tak-  Endoscopist data: full name of endoscopist performing
ing biopsy samples by endoscopists and to analyze the as- the procedure and assisting physician (if present).
sociation between a novel quality indicator, the All indications for EGD were coded and assigned into 4
endoscopist biopsy rate (EBR), and detection of GPCs in groups: (1) symptom evaluation (1a, benign symptoms; 1b,
the stomach and the rate of missed GCs. alarm symptoms); (2) surveillance of premalignant conditions;
(3) cancer and non-epithelial neoplasms (known cancer, high
suspicion of cancer in imaging tests, cancer during treatment,
METHODS cancer follow-up); and (4) others. We linked EGDs with corre-
sponding histopathology reports. Each histopathology finding
Study design was assigned to the esophagus, stomach, or duodenum and
This was a multicenter, retrospective, cohort study then characterized as normal mucosa (no pathology in the
analyzing outpatient EGD databases and histopathology re- microscopic examination), non-neoplastic lesions (eg, acute
ports from 2 high-volume, geographically distinct , endos- gastritis, fundic gland polyp), precancerous condition (eg,
copy units between 2002 and 2015: Unit A, the endoscopy atrophic gastritis, intestinal metaplasia), cancer, or non-
unit at the Department of Cancer Prevention, The Maria epithelial neoplasm (eg, neuroendocrine tumors, gastrointes-
Sklodowska-Curie Memorial Cancer Center and Institute tinal stromal tumors, lymphomas).
of Oncology, Warsaw; Unit B, the endoscopy unit at the For each endoscopist, we counted EBR values and then
Department of Gastroenterology, Pomeranian Medical Uni- divided them by the quartile values into 4 groups, corre-
versity, Szczecin, Poland. sponding to low, moderate, high, and very-high EBR. We
The database from Unit A was used as a derivation cohort analyzed the association between EBR groups and detec-
to assess the EBR parameter and to analyze its correlation tion of GPCs (atrophic gastritis, intestinal metaplasia, and
with GPC detection. Unit B database was used for external low-, indefinite-, and high-grade dysplasia), and detection
validation. Both databases were used to find an association of all UGI premalignant conditions, including GPCs, but
between EBR and the rate of missed cancers (detailed meth- also Barrett’s esophagus (BE), squamous intraepithelial
odology explained below). The research proposal was re- neoplasm (SIN), and duodenal adenomas (DAs).

2 GASTROINTESTINAL ENDOSCOPY Volume -, No. - : 2019 www.giejournal.org

Januszewicz et al Biopsy rate as a gastroscopy quality marker

Derivation cohort Validation cohort

17,490 EGDs 14,589 EGDs

2,150 EGDs 295 EGDs

32 pts. < 18yrs.
26 pts. < 18yrs.
2,051 by trainees
269 missing data
67 missing data

15,340 EGDs 14,294 EGDs

Low Moderate High Very high Low Moderate High Very high
EBR EBR EBR EBR EBR EBR EBR EBR

5,172 2,555 3,666 3,947 10,003 1,785 2,387 119

EGDs EGDs EGDs EGDs EGDs EGDs EGDs EGDs

National Cancer Registry

288 baseline GCs 36 missed GCs* 26 latent GCs¶

*, > 1 month AND < 3 years Low EBR High EBR

¶, > 3 years
GC diagnosis after clear EGD Moderate EBR Very high EBR

29 missed GCs 7 missed GCs

Figure 1. Study flowchart. EBR, Endoscopists biopsy rate; GC, gastric cancer.

Validation group (Unit B) assumption was made that cancers diagnosed within 3
The Unit B database constituted the validation cohort. For years after negative gastroscopy were already present at
this database, we used the same inclusion and exclusion the time of the initial examination as either early
criteria as for the derivation cohort and coded the endoscopy malignancy or precursor lesions. Cancers diagnosed at the
reports and histopathology findings in a similar fashion. How- initial EGD (and within 1 month) were considered
ever, this database did not contain information on indications baseline cancers (we decided that 1 month should include
for EGD, type of the endoscope, and sedation. all cases of delayed histopathology results), and those
diagnosed after 3 years were classified as latent cancers.
Missed cancers
We followed up all patients from Unit A and Unit B using Statistical methods
their personal identification numbers (PESEL) through the Baseline characteristics were prepared using medians,
National Cancer Registry to identify those diagnosed with interquartile ranges, and contingency tables. Spearman
GC coded C.16 according to the International Classification correlation was used to measure the association between
of Diseases, 10th Revision – ICD 10 before December 31, EBR and the risk of GPC and dysplasia at EGD. We analyzed
2015. Missed cancers were defined as those diagnosed after the association between EBR groups and the detection of
1 month and within 3 years after an EGD showing no evi- GPCs and all UGI premalignant conditions using multivar-
dence of cancer. We used the same criteria for missed iate clustered logistic regression models. Because 1 patient
GCs as in several previous studies within the field.4,8-11 may have had more than 1 EGD, standard errors of model
This definition is based on observation studies by Fujita12 estimates were clustered according to patients to adjust for
showing that the doubling time of GC on the mucosal intrapatient correlation. We reported odds ratios (ORs)
surface is approximately 2 to 3 years. Therefore, an and 95% confidence intervals (CIs) adjusted for the

www.giejournal.org Volume -, No. - : 2019 GASTROINTESTINAL ENDOSCOPY 3

Biopsy rate as a gastroscopy quality marker Januszewicz et al

TABLE 1. Characteristics of the derivation cohort (Unit A)

Atrophic Intestinal Gastric Gastric

EGDs, n (%) All GPC, n (%)* gastritis, n (%) metaplasia, n (%) dysplasia, n (%) cancer, n (%)

All patients 15,340 (100) 1331 (8.7) 318 (2.1) 825 (5.4) 188 (1.2) 188 (1.2)
Age groups (median age, 57 years; range: 18- 91 years)
18-49 years 4371 (28.5) 174 (1.1) 53 (0.3) 100 (0.7) 21 (0.1) 25 (0.2)
50-69 years 7992 (52.0) 727 (4.7) 168 (1.1) 459 (3.0) 100 (0.6) 106 (0.7)

70 years 2997 (19.5) 430 (2.8) 97 (0.6) 266 (1.7) 67 (0.4) 57 (0.3)
Gender
Female 8976 (58.5) 728 (4.7) 197 (1.3) 452 (2.9) 79 (0.5) 63 (0.4)
Male 6364 (41.5) 603 (3.9) 121 (0.8) 373 (2.4) 109 (0.7) 125 (0.8)
Indications
Benign symptomsy 7766 (50.6) 516 (3.3) 147 (0.9) 321 (2.1) 48 (0.3) 21 (0.1)
Alarm symptomsz 1691 (11.0) 190 (1.2) 67 (0.4) 109 (0.7) 14 (0.1) 28 (0.2)
Others 5883 (38.4) 625 (4.1) 104 (0.7) 395 (2.6) 126 (0.8) 139 (0.9)
GPC, Gastric precancerous condition.
*Atrophic gastritis, intestinal metaplasia, and gastric dysplasia.
yDyspepsia, abdominal pain, and reflux.
zAnemia, GI bleeding, dysphagia, weight loss, emesis.

patient’s gender and age at the time of diagnosis. All statis- were assigned to each group (4 in each group). The ORs
tical tests were 2-sided. A P value of less than .05 was of the moderate, high, and very-high EBR groups in detect-
considered to indicate statistical significance. All analyses ing GPC were 1.6 (95% CI, 1.3-1.9; P < .001), 2.0 (95% CI,
were performed with Stata software, version 13.1 (Stata 1.7-2.4; P < .001), and 2.5 (95% CI, 2.1-2.9; P < .001),
Corporation, College Station, Tex). respectively, compared with the low EBR group. We per-
formed additional analysis extracting only detection of
GD (any grade). For GD, the ORs were 2.1 (95% CI, 1.1-
RESULTS 4.0; P Z .024), 2.7 (95% CI, 1.5-4.7; P Z .001), and 1.9
(95% CI, 1.1-3.4; P Z .025), for the moderate, high, and
Derivation cohort very-high EBR group, respectively, compared with the
Unit A database consisted of 17,490 EGD reports on low EBR group. For all UGI premalignant conditions, the
13,875 patients, of which 2150 (12.3%) met the exclusion ORs for the moderate, high, and very-high EBR groups
criteria; the remaining 15,340 (87.7%) were analyzed were 1.6 (95% CI, 1.3-1.9; P < .001), 1.9 (95% CI, 1.6-2.2;
(12,433 patients). The study flowchart is presented in P < .001), and 2.5 (95% CI, 2.1-2.9; P < .001), respectively,
Figure 1. Most patients (10,503; 84.5%) had 1 EGD, 1365 compared with the low EBR group.
(11.0%) had 2 EGDs, and 565 (4.5%) had 3 or more In addition, with regard to the indications for the EGDs,
EGDs (maximum 14). In total, 1331 (12.7%) GPCs were we performed sensitivity analysis by excluding all proced-
detected, including 188 diagnoses of gastric dysplasia ures with indications of previously recognized premalig-
(GD) (14.2%; any grade), 150 BE (1.0%), 16 SINs (0.1%), nant conditions, genetic cancer syndromes, cancer or
and 49 DAs (0.3%). The median patient age was 57 years high suspicion of cancer, and non-epithelial neoplasms.
(range, 18-91 years), and the most common indication This excluded 4690 EGDs (26.8%); the remaining 12,800
for EGD was upper abdominal pain, accounting for 4379 EGDs were analyzed. The association between EBR and
procedures (28.5%). Most of the EGDs (n Z 8976) were GPC detection was maintained. The ORs for the moderate,
performed on female patients (58.5%). Nearly all high, and very-high EBR group for detecting GPC were 1.8
procedures were performed with local anesthetic only (95% CI, 1.4-2.2; P < .001), 2.2 (95% CI, 1.8-2.7; P < .001),
(n Z 15,317, 99.8%) and with the use of standard video- and 3.2 (95% CI, 2.7-3.8; P < .001), and for GD detection,
endoscopes (93.6%). Detailed characteristics of the Unit The ORs were 2.3 (95% CI, 1.1-4.8; P Z .021), 2.8 (95% CI,
A cohort are presented in Table 1. 1.5-5.2; P Z .002), and 2.1 (95% CI, 1.1-3.9; P Z .026),
The EBR value of 16 endoscopists in this cohort varied respectively, compared with the low EBR group.
between 22.4% and 65.8%. The median EBR value was
43.7%, and the quartile values of EBR were as follows: Validation cohort
22.4% to 36.7% (low EBR group), 36.8% to 43.7% (moder- Unit B database consisted of 14,589 EGD reports on
ate EBR group), 43.8% to 51.6% (high EBR group), and 11,333 patients, of which 295 (2.0%) met the exclusion
51.7% to 65.8% (very-high EBR group). Endoscopists criteria; the remaining 14,294 (98.0%) were analyzed. Most

4 GASTROINTESTINAL ENDOSCOPY Volume -, No. - : 2019 www.giejournal.org

Januszewicz et al Biopsy rate as a gastroscopy quality marker

TABLE 2. Characteristics of the endoscopists and EBR group for the derivation and validation cohorts (Unit A and Unit B)

EBR group
Endoscopist EBR (%) EGDs, n GPC, n (%)* GD, n (%) (EBR range) OR for GPC 95% CI OR for GD 95% CI

Derivation cohort (EGD, n[15,340)

1 22.4 1700 41 (2.4) 6 (0.3) Low EBR (22.4-36.7) 1.0 - 1.0 -
2 23.4 470 25 (5.3) 2 (0.4)
3 28.4 264 10 (3.8) 1 (0.4)
4 35.5 2738 197 (7.2) 29 (0.5)
5 37.9 140 8 (5.7) 0 (0.0) Moderate EBR (36.8-43.7) 1.6 1.3-1.9 2.1 1.1-3.9
6 40.5 1714 131 (7.6) 24 (0.6)
7 40.9 428 44 (10.3) 14 (1.2)
8 43.2 273 20 (7.3) 4 (1.1)
9 44.1 290 24 (8.3) 3 (0.7) High EBR (43.8-51.6) 2.0 1.7-2.4 2.7 1.5-4.7
10 46.1 1682 139 (8.3) 24 (0.6)
11 49.1 422 49 (11.6) 9 (1.4)
12 50.2 1272 156 (12.3) 23 (1.4)
13 53.0 415 51 (12.3) 4 (0.5) Very-high EBR (51.7-65.8) 2.5 2.1-2.9 1.9 1.1-3.4
14 55.3 517 57 (11.0) 4 (0.4)
15 58.6 278 30 (10.8) 2 (0.7)
16 65.8 2737 349 (12.8) 39 (0.8)
Validation cohort (EGD n[14,294)
A 22.0 413 0 (0.0) 0 (0.0) Low EBR (22.0-36.7) 1.0 - 1.0 -
B 22.4 339 3 (0.9) 3 (0.9)
C 27.4 3289 52 (1.6) 17 (0.5)
D 30.4 1,656 47 (2.8) 26 (1.6)
E 30.5 617 16 (2.6) 4 (0.6)
F 35.4 3689 82 (2.2) 31 (0.8)
G 41.2 1338 21 (1.6) 9 (0.7) Moderate EBR (36.8-43.7) 0.8 0.5-1.2 0.9 0.5-1.6
H 42.1 447 6 (1.3) 3 (0.7)
I 47.1 2387 43 (1.8) 21 (0.9) High EBR (43.8-51.6) 3.0 2.4-3.7 1.0 0.6-1.6
J 52.9 119 15 (12.6) 4 (3.4) Very-high EBR (
51.7) 5.6 3.2-9.8 3.8 1.3-10.6
EBR, Endoscopist biopsy rate; GPC, gastric precancerous conditions; GD, gastric dysplasia; OR, odds ratio; CI, confidence interval.
*Including atrophic gastritis, intestinal metaplasia, low-, indefinite-, and high- grade dysplasia.

patients (9425; 83.2%) had 1 EGD, 1334 (11.8%) had 2 Overall, for Units A and B combined, the endoscopist
EGDs, and 574 (5.1%) had 3 or more EGDs (maximum EBR value was strongly correlated with detection of GPC
14). In total, 285 GPCs were detected, including 118 (r Z 0.83; P < .001) and a trend was seen for GD detec-
(0.8%) GD (any grade). The median patient age was 56 years tion, however without statistical significance (r Z 0.39;
(range, 18-96 years). As in the derivation cohort, most of the PZ.057) (Fig. 2).
EGDs (n Z 8932) were performed on females (62.5%).
The EBR range of 10 endoscopists in Unit B varied be- EBR and missed gastric cancers
tween 22.0% and 52.9%. Using the EBR quartile ranges es- For this analysis, we divided all endoscopists based on
tablished in Unit A, we assigned endoscopists from Unit B the median EBR value into higher (
43.8%) and lower
into 4 EBR groups and analyzed the ORs of GPC detection. (<43.8%) EBR groups to analyze the association between
The ORs for the moderate, high, and very-high EBR groups EBR group and the odds of missed cancer during EGD.
for detecting GPCs were 0.8 (95% CI, 0.5-1.2), 3.0 (95% CI, A total of 350 GCs were diagnosed at Units A and B in the
2.4-3.7), and 5.6 (95% CI, 3.2-9.8), respectively, compared study period. This included 288 GCs identified at baseline
with the low EBR group. For detecting GD, the ORs endoscopy and 62 GCs identified through cross-reference
were 0.9 (95% CI, 0.5-1.6), 1.0 (95% CI, 0.6-1.6), and 3.8 with the National Cancer Registry. Of the latter, 36 GCs
(95% CI, 1.3-10.6), compared with the low EBR group. were classified as missed cancers (18 in Unit A and 18 in
EBR values from Units A and B are presented in Table 2. Unit B; median time from EGD to cancer diagnosis, 0.9

www.giejournal.org Volume -, No. - : 2019 GASTROINTESTINAL ENDOSCOPY 5

Biopsy rate as a gastroscopy quality marker Januszewicz et al

16%

rs=0.835, P< .001

14%

12%

Premalignant conditions (Unit A)

10%
Endoscopic detection

Premalignant conditions (Unit B)

8%

Gastric dysplasia (Unit A)

6%

Gastric dysplasia (Unit B)

4%

2%
rs=0.378, P= .057

0%
0 10 20 30 40 50 60 70 80
Endoscopists’ biopsy rate (EBR)

Figure 2. Combined data from Unit A and Unit B showing the association between endoscopist biopsy rate and detection of gastric premalignant con-
ditions (red circles) and dysplasia (blue circles). Each circle represents a single endoscopists’ performance and the diameter of the circle corresponds to
the number of endoscopies performed by the endoscopist. rs, Spearman correlation.

years; interquartile range, 0.4-1.7 years) and 26 as latent can- microscopic assessment among all EGDs performed. This
cers (median time from EGD to cancer diagnosis, 7.3 years; rate varied among endoscopists between 3.5% and
interquartile range, 4.6-8.5 years) (Fig. 1). The overall GC 41.0%. The low EBR group had the lowest rate of negative
miss rate was 10.3%. Twenty-nine of the missed cancers biopsy results (mean, 10.8%; range, 3.5%-20.5%) compared
were diagnosed by endoscopists with a lower EBR with the other groups because the mean rates of negative
(80.6%), and 7 by those with a higher EBR (19.4%). biopsy results for the moderate, high, and very-high EBR
Using a logistic regression model adjusted for patient groups were 20.5% (range, 8.8%-30.0%; P Z .015), 23.3%
age, gender, and endoscopy unit, we have shown that pa- (range, 5.8%-29.0%; P Z .007), and 32.2% (range, 14.3%-
tients examined by endoscopists with a higher EBR had a 41.0%; P < .001), respectively (Fig. 3).
56% lower risk of missed cancer during EGD compared
with the lower EBR group (OR, 0.44; 95% CI, 0.20-1.00;
P Z .049). The incidence of missed GC in the lower EBR DISCUSSION
group was 49.6 per 100,000 person-years and in the higher
EBR group, it was 23.1 per 100,000 person-years. The risk Unlike in the field of colonoscopy, for which multiple
difference was 26.5 per 100,000 person-years (95% CI, 1.7- quality indicators have been identified,13-15 there are few
51.4 per 100,000 person-years). performance measures for EGD, and most of them are
not validated, are based on low-quality data, and are rarely
EBR and the rate of negative biopsy results associated with patient-oriented outcomes. In our study,
For all endoscopists included in the study, we analyzed we have characterized and validated a new performance
the rate of negative biopsy results, defined as the propor- measure, EBR, based on a hypothesis that the rate of ob-
tion of biopsy samples showing no abnormality in the taining biopsy samples, similar to the polyp detection

6 GASTROINTESTINAL ENDOSCOPY Volume -, No. - : 2019 www.giejournal.org

Januszewicz et al Biopsy rate as a gastroscopy quality marker

The rate of negative biopsies within EBR groups

35

30

25

20

15

10

0
Low EBR Moderate EBR High EBR Very-high EBR
(22.4%-36.7%) (36.8%-43.7%) (43.8%-51.6%) (51.7%-65.8%)

Figure 3. The rate of negative biopsy results within the EBR groups. EBR, Endoscopist biopsy rate.

rate in colonoscopy, broadly reflects the number of abnor- those in our study (22.0%-65.8%); however, the South
malities detected in routine outpatient EGDs. We have Korean study involved an asymptomatic screening
found that the biopsy rate was markedly variable among population, whereas our cohort included patients being
endoscopists, and this was observed both in the derivation evaluated for GI symptoms, who more often require
cohort (EBR range, 22.4%-65.8%) and the external valida- biopsy sampling. In comparison, in a previous Japanese
tion cohort (EBR range, 22.0%-52.9%). We have shown study on symptomatic patients, the mean biopsy rate was
that the EBR parameter was strongly associated with GPC 55.0%, which was similar to ours; however, the higher rate
detection (r Z 0.83; P < .001) and, most importantly, of GC in Japan and widespread use of advanced imaging
with the risk of missed GCs. The decision to use these end- techniques are confounders in this comparison.8
points was supported by the fact that GC is the most com- By cross-linking our data with the National Cancer Reg-
mon UGI malignancy in Western countries, and GPC istry, we investigated the rate of missed GCs. We identified
detection was previously used in other studies on UGI 36 EGDs with negative results for cancer in patients who
quality indicators.17 Nevertheless, the usefulness of EBR were diagnosed with GC between 1 month and 3 years af-
was also maintained by including all UGI premalignant terward. We used these criteria in accordance with previ-
conditions (GPCs, BE, SIN, DA) in the analysis. ous studies analyzing the rate of missed upper GI
Recently, several gastroenterology societies have pub- cancers.4,8-11 The GC miss rate in our cohort was 10.3%.
lished guidelines and statement papers on performance In comparison, Raftopoulos et al9 reported a missed
measures for UGI endoscopy.5-7 To date, the most broadly cancer rate of 6.7% in a cohort of 28,000 patients.
studied parameter for EGD is the procedure time. For However, this number included all UGI cancers
example, an association between the examination time for (duodenum, stomach, and esophagus). When including
BE and the detection rate of high-grade dysplasia and adeno- only GCs, the missed cancer rate was 4.3%, which is over
carcinoma was shown.16 In a subsequent study by Teh 2-fold lower than in our study. On the other hand,
et al,17 endoscopists with a mean EGD examination time previous Asian studies reported a missed GC rate of 14%
of more than 7 minutes were more likely to detect and 26%,8,10 which is substantially higher than in our study.
precancerous lesions and cancers in the UGI tract However, these reports originate from countries with a
compared with those with a shorter examination time. high incidence of GCs, with different histologic criteria
Finally, a recent study by Park et al18 found that for cancer.
endoscopists with a mean examination time of more than The EBR parameter was assessed using a large dataset of
3 minutes (withdrawal time after reaching the duodenum nearly 30,000 EGDs performed by 26 endoscopists. The
and cleaning the gastric mucosa) were more likely to main strength of our study is that it includes external vali-
detect gastric adenomas and cancers than “fast dation in a high-volume, geographically distinct unit. The
endoscopists.” In this study, in relation to ours, the EBR parameter is easy to calculate because it only requires
frequency of endoscopic biopsies varied significantly the number of EGDs with at least 1 biopsy sample obtained
among endoscopists (range, 6.9%-27.8%) and was strongly from any part of the UGI tract (this can be determined by
correlated with the rate of neoplasm detection (R2 Z 0.76; ICD-9 coding) and the total number of procedures. In the
P Z .015)18 The biopsy rates were notably lower than EBR calculations, we included EGDs with biopsy samples

www.giejournal.org Volume -, No. - : 2019 GASTROINTESTINAL ENDOSCOPY 7

Biopsy rate as a gastroscopy quality marker Januszewicz et al

obtained from any part of the UGI tract (esophagus, stom- The usefulness of EBR requires further evaluation in
ach, duodenum) to make the EBR calculation as simple as prospective trials to determine the most accurate range
possible. We have shown that EBR is associated with the of its value that would represent the highest diagnostic
detection of all UGI premalignant conditions; however, in yield and the lowest cost burden. Increasing EBR values
presenting our results we have focused on gastric findings are associated with an increasing number of negative bi-
(GPC detection and rate of missed GCs) as a surrogate opsy results, and hence costs, and our study may suggest
endpoint to objectively compare endoscopist perfor- that endoscopists within the high EBR group (43.8%-
mance, as was done in previous studies.17 51.6%) represent the best balance between UGI high-risk
Our study was limited by a few relevant factors, and a lesion detection and missed GC diagnoses and the rate
retrospective design is the main one. The EBR parameter of negative biopsy results (costs).
was developed in the setting of routine outpatient endos- However, we are aware that the EBR parameter is depen-
copy units and is not applicable in the inpatient setting, dent on the regional prevalence of GPC and GC. Therefore,
where therapeutic procedures constitute a significant pro- rather as an absolute value, EBR should be used to compare
portion of examinations. Units A and B differed signifi- endoscopists within the same unit/region to see the varia-
cantly in terms of GPC detection, and this can be tion in its value and to identify endoscopists requiring
explained by the difference in expertise: Unit A is an endos- improvement. In our view, EBR indirectly informs about
copy unit based around an oncology center and Unit B is the quality of inspection of the gastric mucosa (like proced-
based around a general gastroenterology department. ure time). Meticulous inspection of the mucosa translates
Despite those differences in performance, we could still into better endoscopic recognition of GPCs, which need
show a meaningful difference in performance between to be confirmed with a biopsy. The higher rate of GPCs diag-
the EBR groups in the validation cohort, which proves noses is also of value to the patients, who are then triaged to
the reproducibility of this parameter. Our electronic data- the population with increased GC risk in the future.
base did not include information on the patients’ risk fac- In conclusion, EBR is highly variable among endoscop-
tors, such as smoking, medications, body mass index, and sists, and we believe that within a group of comparably
the use of advanced imaging techniques, such as virtual experienced endoscopists, it correlates with a detection
chromoendoscopy. Most of the procedures were per- rate of mucosal abnormalities in the UGI tract during
formed using standard video-resolution endoscopes, EGD. This is the first study to describe and analyze EBR
which represents the standard of care in routine outpatient as a quality indicator for routine diagnostic outpatient
endoscopy practice within the timeframe of the study. On EGD and to show its correlation with important, patient-
one hand, this has the advantage of uniform (hence com- oriented outcomes, such as GPC detection and the rate
parable) equipment being used by the endoscopists in our of missed GCs.
study, but, on the other hand, it does not represent the
current shift in standard of care toward high-definition
endoscopy. Moreover, the procedure time was not re- ACKNOWLEDGMENTS
corded in our databases. This limitation is important
because we believe that EBR and procedure time might The first author would like to acknowledge Dr Massimi-
be correlated and both of these parameters represent liano di Pietro for his valuable remarks during the prepara-
the quality of mucosal inspection during EGD. Finally, we tion of this manuscript.
could not extract data on the distribution and extension
of the GPCs. This limitation is particularly important in
terms of atrophic gastritis, which is associated with signifi-
cant risk of cancerous change when diffused or multifocal, REFERENCES
but with low risk (not requiring surveillance) when only
1. International Agency for Research on Cancer. GLOBOCAN 2012: Esti-
limited to the gastric antrum.19 We also did not have mated Cancer Incidence, Mortality and Prevalence Worldwide in
data on whether biopsy protocols, such as the Sydney 2012. Available at: http://globocan.iarc.fr/. Accessed January 6, 2018.
protocol,20 were followed in both units and could not 2. Global Burden of Disease Cancer Collaboration; Fitzmaurice C, Allen C,
differentiate whether biopsy sampling was targeted or Barber RM, et al. Global, regional, and national cancer incidence, mor-
random. We are also aware that, just like the polyp tality, years of life lost, years lived with disability, and disability-
adjusted life-years for 32 cancer groups, 1990 to 2015: a systematic
detection rate in colonoscopy, the EBR parameter might analysis for the global burden of disease study. JAMA Oncol 2017;3:
be susceptible to “gaming.”21 Knowing about the EBR 524-48.
parameter being monitored may have encouraged 3. Correa P. Human gastric carcinogenesis: a multistep and multifactorial
endoscopists to take more random biopsy samples to processdfirst American Cancer Society award lecture on cancer epide-
boost their EBR value without influencing the patient miology and prevention. Cancer Res 1992;52:6735-40.
4. Menon S, Trudgill N. How commonly is upper gastrointestinal cancer
outcome. A potential solution to this problem could be missed at endoscopy? A meta-analysis. Endosc Int Open 2014;2:E46-50.
to adjust the EBR parameter for the rate of GPC 5. Park WG, Shaheen NJ, Cohen J, et al. Quality indicators for EGD. Am J
diagnosis for each endoscopist. Gastroenterol 2015;110:60-71.

8 GASTROINTESTINAL ENDOSCOPY Volume -, No. - : 2019 www.giejournal.org

Januszewicz et al Biopsy rate as a gastroscopy quality marker

6. Bisschops R, Areia M, Coron E, et al. Performance measures for upper 15. Rex DK, Petrini JL, Baron TH, et al. Quality indicators for colonoscopy.
gastrointestinal endoscopy: a European Society of Gastrointestinal Endos- Am J Gastroenterol 2006;101:873-85.
copy (ESGE) Quality Improvement Initiative. Endoscopy 2016;48:843-64. 16. Gupta N, Gaddam S, Wani SB, et al. Longer inspection time is associ-
7. Beg S, Ragunath K, Wyman A, et al. Quality standards in upper gastro- ated with increased detection of high-grade dysplasia and esophageal
intestinal endoscopy: a position statement of the British Society of adenocarcinoma in Barrett’s esophagus. Gastrointest Endosc 2012;76:
Gastroenterology (BSG) and Association of Upper Gastrointestinal Sur- 531-8.
geons of Great Britain and Ireland (AUGIS). Gut 2017;66:1886-99. 17. Teh JL, Tan JR, Lau LJ, et al. Longer examination time improves detec-
8. Hosokawa O, Tsuda S, Kidani E, et al. Diagnosis of gastric cancer up to tion of gastric cancer during diagnostic upper gastrointestinal endos-
three years after negative upper gastrointestinal endoscopy. Endos- copy. Clin Gastroenterol Hepatol 2015;13:480-7.
copy 1998;30:669-74. 18. Park JM, Huo SM, Lee HH, et al. Longer observation time increases pro-
9. Raftopoulos SC, Segarajasingam DS, Burke V, et al. A cohort study of portion of neoplasms detected by esophagogastroduodenoscopy.
missed and new cancers after esophagogastroduodenoscopy. Am J Gastroenterology 2017;153:460-9.
Gastroenterol 2010;105:1292-7. 19. Dinis-Ribeiro M, Areia M, de Vries AC, et al. Management of precan-
10. Yalamarthi S, Witherspoon P, McCole D, et al. Missed diagnoses in pa- cerous conditions and lesions in the stomach (MAPS): guideline
tients with upper gastrointestinal cancers. Endoscopy 2004;36:874-9. from the European Society of Gastrointestinal Endoscopy (ESGE), Eu-
11. Chadwick G, Groene O, Riley S, et al. Gastric cancers missed during ropean Helicobacter Study Group (EHSG), European Society of Pa-
endoscopy in England. Clin Gastroenterol Hepatol 2015;13:1264-70. thology (ESP), and the Sociedade Portuguesa de Endoscopia
12. Fujita S. Biology of early gastric carcinoma. Pathol Res Pract 1978;163:297-309. Digestiva (SPED). Endoscopy 2012;44:74-94.
13. Kaminski MF, Thomas-Gibson S, Bugajski M, et al. Performance mea- 20. Dixon MF, Genta RM, Yardley JH, et al. Classification and grading of
sures for lower gastrointestinal endoscopy: a European Society of gastritis. The updated Sydney System. International Workshop on
Gastrointestinal Endoscopy (ESGE) quality improvement initiative. the Histopathology of Gastritis, Houston 1994. Am J Surg Pathol
Endoscopy 2017;49:378-97. 1994;20:1161-81.
14. Rees CJ, Thomas Gibson S, Rutter MD, et al. UK key performance indicators 21. Anderson JC, Butterly LF. Colonoscopy: quality indicators. Clin Transl
and quality assurance standards for colonoscopy. Gut 2016;65:1923-9. Gastroenterol 2015;6:e77.

www.giejournal.org Volume -, No. - : 2019 GASTROINTESTINAL ENDOSCOPY 9