Optimal Sequencing in Same-Day Bidirectional Endoscopy: A Tertiary US Healthcare Center Experience

Abstract

Background

Same-day performance of esophagogastroduodenoscopy (EGD) and colonoscopy is called bidirectional endoscopy (BDE). BDE is commonly performed, but the optimal sequence for which procedure to do first is not well established. This is the first study in the US to investigate the optimal sequence for BDE.

Methods

We performed a cohort study of patients with same-day BDE (2003–2018) at our institution. The two study groups were (1) EGD followed by colonoscopy (UL) and (2) Colonoscopy followed by EGD (LU). Endpoints included procedure duration, extent reached, sedation, and diagnostic yield.

Results

22,905 patients underwent BDE, with complete data available for 16,538. 14,325 underwent UL, and 2213 underwent LU. For diagnostic indications, the LU group was more likely to require opiates (OR 2.8, CI 2.5–3.3, p < 0.001), benzodiazepines (OR 3.0, CI 2.6–3.4, p < 0.001), antihistamines (OR 1.5, CI 1.2–1.8, p < 0.001), while less likely to require anesthesia (OR 0.33, CI 0.28–0.39, p < 0.001). Similar results were found for surveillance indication of endoscopy and in screening colonoscopy, with no differences in EGD screening indications. Total Procedure duration: UL had shorter total procedure duration (23.2 vs. 28.8 min; p < 0.001), EGD duration (5.4 vs. 6.5 min; p < 0.001), and colonoscopy duration (17.9 vs. 22.3 min; p < 0.001). Diagnostic yield: UL and LU sequences had comparable polyp detection rates (37.6% vs. 38.1%, p = 0.65) with similar adenoma detection rates (ADR).

Conclusion

Performing EGD first resulted in shorter procedure times and a reduced likelihood of requiring opioids and benzodiazepines, with no differences regarding procedure-related adverse events or diagnostic yield.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10620-025-08889-0.

Introduction

Esophagogastroduodenoscopy (EGD) and colonoscopy are among the most common gastrointestinal (GI) procedures performed by gastroenterologists for the investigation and management of digestive diseases. Many gastroenterologists have found it beneficial to perform both EGD and colonoscopy sequentially during the same session on the same day, which is termed bidirectional endoscopy (BDE). The practice of BDE has a great potential to deliver high-value care in GI practice, because performing these two procedures on different days involves the addition of extra time, pre-procedural preparation, health care resources (including sedation protocols, utilization of the endoscopy unit, and reimbursement), and most importantly exposing the patient to potential unnecessary procedure-related risks or adverse events [1]. Approximately 10% of endoscopy procedures in the United States are performed as BDE; however, this figure rises to around 65% for patients who undergo both EGD and colonoscopy within a 180-day interval [2]. Same-day BDE can be beneficial, especially when the diagnostic evaluation might require both procedures, such as identifying the bleeding source in an anemic patient. In patients with GI bleeding, undergoing BDE, the prevalence of concurrent upper and lower GI findings has been reported to be as high as 29% [3]. It has also been shown that the same-day BDE reduces sedation usage, length of stay in the hospital, and medical costs [4–6].

Though there are several compelling reasons to perform the same-day BDE, there is little evidence on the best sequence to use. Performing the EGD first can potentially lead to severe bloating of the bowels and difficult intubation during the colonoscopy due to gas insufflation during the EGD. On the other hand, it was thought that undergoing EGD after a colonoscopy was more stressful in non-sedated patients [7, 8]. However, Hsian et al. conducted a randomized controlled trial (RCT) involving 176 patients under propofol-induced sedation, which suggested that BDE was better tolerated and required lower doses of sedation when EGD was performed before colonoscopy [9]. Interestingly, a subsequent RCT with 163 patients found no difference in tolerability and in total dose of sedation between either sequence when moderate sedation was used [10]. Chen et al. performed another RCT with 120 patients that demonstrated EGD first was the best sequence when moderate sedation and carbon dioxide insufflation are used for all endoscopies due to lower recovery team and lower total dose of fentanyl and midazolam that were used in the study [11]. Notably, the above-mentioned studies were conducted outside the United States of America (U.S.), and none investigated the sequence of BDE in a U.S. healthcare setting [9–11]. Thus, our aim was to identify the optimal BDE sequence in a tertiary US healthcare center.

Methods

We conducted a retrospective cohort study examining patients who underwent same-day BDE from 2003 to 2018 at our institution. The study compared two groups: patients who had UL vs. LU. We collected data on demographic characteristics, procedure type, duration, sedation doses, findings, and pathology results.

Study Setting

Data were sourced from the institution’s prospectively maintained endoscopy registry using Provation endoscopy software. The Cleveland Clinic’s institutional review board approved the study.

Subjects

The study included patients who underwent BDE under moderate conscious sedation, monitored anesthesia care (MAC), or general anesthesia (GA), with MAC and GA combined into an “anesthesia subgroup” for analysis. Patients were excluded if critical variables like scope-in/out times or extent reached during EGD or colonoscopy were missing. Only the first recorded BDE session was included for patients who had multiple procedures.

Outcomes

The primary outcomes measured were procedure duration, extent reached, sedation doses, adverse events, and diagnostic yield. The duration of upper and lower endoscopic procedures was defined as the time from scope insertion to scope withdrawal. The time taken to rotate the bed, and pharyngeal anesthesia for EGD, during the procedure is not included in the final calculation. This information is recorded by the endoscopy nurses in the electronic health record system. Satisfactory extent was defined as reaching the duodenum’s second portion for EGD and the cecum for colonoscopy. We also calculated the time to reach the cecum for colonoscopy. Sedation doses were recorded and categorized by medication type, including opioids (fentanyl and meperidine) and benzodiazepines (midazolam), with meperidine doses converted to fentanyl equivalents. Additional sedation with IV diphenhydramine or diazepam, when documented, was included as “extra enhancing-sedation medications.”

Diagnostic Yield

Procedures were categorized by indication, including screening, surveillance, and diagnostic EGD, as well as screening, surveillance, diagnostic colonoscopy, and IBD surveillance. The adenoma detection rate (ADR) for screening colonoscopy was compared between the two groups.

Adverse Events

All documented adverse events noted in procedure reports were included.

Other Variables

Abstracted EGD findings included ulcers, AVMs, esophagitis, Barrett’s esophagus, and other pathologies, while colonoscopy findings included polyps, inflammation, and IBD-related abnormalities. Recorded maneuvers during procedures comprised biopsies, polypectomies, dilations, and other interventions. Demographic data such as age, gender, race, BMI, ASA classification, procedure setting, fellow involvement, and IBD history were also collected.

Statistical Analysis

Data are presented as mean ± standard deviation, median [25th, 75th percentiles] or N (%). A univariate analysis was performed to assess differences between subjects who had UL vs. those who had LU; analysis of variance (ANOVA) or the non-parametric Kruskal–Wallis tests were used to compare continuous or ordinal variables and Pearson’s chi-square tests, or Fisher’s Exact tests were used for categorical factors. In addition, multivariable regression analyses were performed to assess the associations between sequence and outcomes of interest while adjusting for age, gender, ethnicity, patient setting and endoscopy unit location. We also accounted for IBD specialists (assuming that they may prefer performing colonoscopy first) and we adjusted for it, as well. Linear regression was used to model total procedure duration and medication dosages and logistic regression was used to model use of medications. SAS (version 9.4, The SAS Institute, Cary, NC) was used for all analyses and a p < 0.05 was considered statistically significant.

Results

Cohort Description

22,905 patients underwent bidirectional endoscopy, of whom 16,538 had complete data for the study and were included in the analysis. The cohort has a mean age of 59.4 years and 54.3% were females with mean BMI of 29.8 kg/m². UL was performed on 14,325 patients, whereas LU was performed on 2213 patients. There were no significant differences between the two groups regarding sex, age, BMI, or smoking status. The main difference between groups is the higher incidence of IBD in the LU group (3.3% in UL sequence and 5.7% in LU sequence). The major indication for EGD (96.3%) and colonoscopies (61.3%) was diagnostic evaluation. More fellows were involved during the LU sequence compared to UL sequence (22.8% vs. 7.7%, p < 0.001). Most colonoscopies had a good or excellent bowel preparation (79% in UL sequence and 77% in LU sequence) (Table 1).

Table 1.

Demographics and overall procedure characteristics

Category	Upper then lower (n = 14,325)	Lower then upper (n = 2213)	p value
Demographics
Age (years)	56.9 ± 15.7	57.2 ± 15.5	0.37 (a)
Gender	14,268	2210	0.076 (c)
Female	8535 (59.8%)	1278 (57.8%)
Male	5733 (40.2%)	932 (42.2%)
Race	13,729	2138	0.044 (c)
White	11,155 (81.3%)	1690 (79.0%)
Black	2084 (15.2%)	357 (16.7%)
Other	490 (3.6%)	91 (4.3%)
BMI (kg/m2)	29.9 ± 6.2	29.3 ± 7.2	< 0.001 (c)
Smoking	5670 (39.6%)	838 (37.9%)	0.12 (c)
IBD (inflammatory bowel disease)	472 (3.3%)	127 (5.7%)	< 0.001 (c)
Anesthesia/sedation characteristics
Anesthesia use	2905 (20.3%)	254 (11.5%)	< 0.001 (c)
Opiate use	8862 (61.9%)	1843 (83.3%)	< 0.001 (c)
Benzodiazepines use	8944 (62.4%)	1855 (83.8%)	< 0.001 (c)
Antihistamine use	814 (5.7%)	128 (5.8%)	0.85 (c)
EGD
Endoscopy indication	14,325	2213	0.91 (c)
Screening	153 (1.1%)	24 (1.1%)
Surveillance	385 (2.7%)	56 (2.5%)
Diagnostic	13,787 (96.2%)	2133 (96.4%)
Fellow involved during endoscopy	1099 (7.7%)	505 (22.8%)	< 0.001 (c)
Colonoscopy
Colonoscopy indication	14,325	2213	< 0.001 (c)
Screening	2868 (20.0%)	438 (19.8%)
Surveillance	1931 (13.5%)	406 (18.3%)
Diagnostic	9526 (66.5%)	1369 (61.9%)
Colon preparation quality	14,247	2213	< 0.001 (c)
Poor	1004 (7.0%)	137 (6.2%)
Fair	1987 (13.9%)	370 (16.7%)
Good	9492 (66.6%)	1362 (61.5%)
Excellent	1764 (12.4%)	344 (15.5%)
Fellow involved during colonoscopy	1107 (7.7%)	491 (22.2%)	< 0.001 (c)

Statistics presented as mean ± SD or n (column %). p values indicate analysis of variance (a) or Pearson’s Chi-square test (c)

BMI body mass index, IBD inflammatory bowel disease, EGD esophagogastroduodenoscopy

Procedure Duration

The total bidirectional procedure duration was shorter for UL (mean 23.2 vs. 28.8 min, p < 0.001). EGD and colonoscopies were also shorter in UL procedures compared to LU. The EGD meantime was 5.4 ± 3.7 (UL) vs. 6.5 ± 4.0 min (LU), p < 0.001, and the colonoscopy meantime was 17.9 ± 9.4 (UL) vs. 22.3 ± 11.7 min (LU), p < 0.001. In nearly all subgroups, LU procedures consistently took longer than UL procedures, regardless of the procedure’s indication. The only exception, which was not statistically significant, was the surveillance EGD subgroup. The procedures performed by GI fellows (trainees) had longer durations overall. Notably, the difference in procedure time between the UL and LU sequences remained significant even within this group, with LU procedures taking longer (Table 2, Supplementary Fig. 1, Supplementary Fig. 2).

Table 2.

Procedure sequence and total procedure time sub-group analysis results

Sub-group	Adjusted mean total procedure time (min)		p value
	Upper then lower	Lower then upper
Screening EGD	25.7 (21.2, 30.2)	33.8 (27.4, 40.3)	0.002
Surveillance EGD	23.6 (19.0, 28.1)	25.3 (19.7, 30.8)	0.36
Diagnostic EGD	23.8 (20.6, 27.0)	26.9 (23.7, 30.1)	< 0.001
No fellow involved during EGD	22.0 (18.0, 25.9)	24.7 (20.8, 28.7)	< 0.001
Fellow involved during EGD	28.1 (21.7, 34.4)	30.2 (23.7, 36.8)	0.004
Screening colonoscopy	22.1 (14.9, 29.4)	23.5 (16.2, 30.8)	0.026
Surveillance colonoscopy	23.3 (15.7, 31.0)	25.3 (17.6, 33.0)	0.001
Diagnostic colonoscopy	24.4 (20.4, 28.4)	28.1 (24.0, 32.1)	< 0.001
No fellow involved during colonoscopy	22.0 (18.0, 25.9)	24.7 (20.7, 28.6)	< 0.001
Fellow involved during colonoscopy	27.8 (21.4, 34.1)	30.4 (23.9, 36.9)	< 0.001
No IBD	23.9 (20.7, 27.1)	26.6 (23.4, 29.8)	< 0.001
IBD	23.8 (21.1, 26.5)	30.1 (26.7, 33.4)	< 0.001

*Adjusted for age, gender, ethnicity, patient setting, and endoscopy suite location

IBD inflammatory bowel disease, EGD esophagogastroduodenoscopy

Anesthesia/Sedation Characteristics

Anesthesia (general, MAC, and moderate sedation) was more commonly used in the UL sequence compared to the LU sequence (20.3% vs. 11.5%, p < 0.001), with significant differences observed across all sub-analyses, except for screening EGD. Opiate (61.9% in UL vs. 83.3% in LU, p < 0.001) and benzodiazepine (62.4% in UL vs. 83.8% in LU, p < 0.001) use was more prevalent in the LU sequence in all subgroup analyses, again apart from screening EGD. While there were no differences in the general analysis (5.7% vs. 5.8%, p = 0.85), antihistamine use was higher in certain sub-analyses for the UL sequence (Table 3 and Fig. 1).

Table 3.

Procedure sequence and medication use (including anesthesia cases) sub-group analysis results

Subgroup	Opiates (LU vs. UL) OR (95% CI)	p value	Antihistamine (LU vs. UL) OR (95% CI)	p value	Benzodiazepines (LU vs. UL) OR (95% CI)	p value	Anesthesia (LU vs. UL) OR (95% CI)	p value
Screening EGD	2.1 (0.88, 4.8)	0.094	1.6 (0.41, 6.1)	0.51	2.0 (0.76, 4.8)	0.10	0.57 (0.23, 1.5)	0.24
Surveillance EGD	2.8 (2.4, 3.1)	< 0.001	1.2 (1.01, 1.5)	0.04	2.9 (2.5, 3.2)	< 0.001	0.36 (0.31, 0.42)	< 0.001
Diagnostic EGD	2.8 (2.5, 3.3)	< 0.001	1.5 (1.2, 1.8)	< 0.001	3.0 (2.6, 3.4)	< 0.001	0.33 (0.28, 0.39)	< 0.001
No fellow involved during EGD	1.9 (1.4, 2.5)	< 0.001	0.49 (0.26, 0.92)	0.025	1.8 (1.4, 2.4)	< 0.001	0.56 (0.42, 0.74)	< 0.001
Fellow involved during EGD	2.4 (1.4, 2.5)	< 0.001	1.05 (0.61, 1.8)	0.87	2.6 (1.9, 3.5)	< 0.001	0.36 (0.24, 0.53)	< 0.001
Screening colonoscopy	2.5 (1.9, 3.2)	< 0.001	0.64 (0.35, 1.2)	0.15	2.5 (1.9, 3.2)	< 0.001	0.39 (0.29, 0.54)	< 0.001
Surveillance colonoscopy	2.8 (2.4, 3.2)	< 0.001	1.6 (1.2, 2.0)	< 0.001	2.9 (2.4, 3.3)	< 0.001	0.37 (0.31, 0.44)	< 0.001
Diagnostic colonoscopy	2.8 (2.4, 3.2)	< 0.001	1.5 (1.2, 1.8)	< 0.001	2.9 (2.5, 3.3)	< 0.001	0.34 (0.29, 0.40)	< 0.001
No fellow involved during colonoscopy	2.0 (1.5, 2.7)	< 0.001	0.48 (0.26, 0.89)	0.021	2.0 (1.5, 2.6)	< 0.001	0.52 (0.39, 0.71)	< 0.001
Fellow involved during colonoscopy	2.7 (2.4, 3.0)	< 0.001	1.3 (1.02, 1.6)	0.034	2.8 (2.4, 3.1)	< 0.001	0.37 (0.32, 0.43)	< 0.001
No IBD	2.9 (1.8, 4.8)	< 0.001	1.2 (0.48, 3.1)	0.68	2.9 (1.8, 4.7)	< 0.001	0.39 (0.23, 0.64)	< 0.001

All ORs are for lower then upper vs. upper then lower and are adjusted for age, gender, ethnicity, patient setting and endoscopy suite location

IBD inflammatory bowel disease, EGD esophagogastroduodenoscopy, OR odds ratio, CI confidence interval

Fig. 1.

Medication use (unadjusted analysis)

The medication doses (benzodiazepines and opiates) were generally similar between the groups, although certain subgroups exhibited higher doses in the UL sequence, while others did so in the LU sequence (Supplementary Table 2 and Supplementary Table 2).

EGD

The EGD mean time was 5.4 (UL) vs. 6.5 min (LU), p < 0.001. More than 96% of EGD indications are diagnostic, followed by surveillance and screening. When comparing different procedural sequences—UL versus LU—biopsies were more frequently performed in the UL sequence (76.0% vs. 66.6%, p < 0.001), though the Barrett’s biopsy rates remained similar between the two. On the other hand, procedures such as feeding tube placement, Bravo, and video capsule placement were more commonly done in the LU sequence. About the EGD findings: Esophagitis was more frequently detected with UL compared to LU (8.7% vs. 4.0%, p < 0.001). Similarly, hiatal hernia was identified more often with UL (0.23% vs. 0%, p = 0.024) (Table 4).

Table 4.

EGD characteristics

Factor	Upper then lower (n = 14,325)	Statistics (upper then lower)	Lower then upper (n = 2213)	Statistics (lower then upper)	p value
Indication					< 0.001 b
Screening	14,325	153 (1.1%)	2213	24 (1.1%)
Surveillance	14,325	385 (2.7%)	2213	56 (2.5%)
Diagnostic	14,325	13,787(96.2%)	2213	2,133 (96.4%)
Procedure duration (min)	14,325	5.4 ± 3.7	2213	6.5 ± 4.0	< 0.001a
EGD maneuvers
Biopsy	14,325	11,022(76.9%)	2213	1,473 (66.6%)	< 0.001b
Barrett’s biopsy	11,022	467 (4.2%)	1473	63 (4.3%)	0.94b
Dilation	14,325	471 (3.3%)	2213	73 (3.3%)	0.98b
Clips	14,325	85 (0.59%)	2213	13 (0.59%)	0.97b
Polypectomy	14,325	521 (3.6%)	2213	86 (3.9%)	0.56b
Thermal therapy	14,325	152 (1.1%)	2213	24 (1.1%)	0.92b
Banding	14,325	75 (0.52%)	2213	15 (0.68%)	0.36b
Feeding tube placement	14,325	15 (0.10%)	2213	8 (0.36%)	0.003b
Bravo placement	14,325	16 (0.11%)	2213	30 (1.4%)	< 0.001b
Video capsule placement	14,325	4 (0.03%)	2213	12 (0.54%)	< 0.001b
EGD findings
Ulcer	14,325	479 (3.3%)	2213	59 (2.7%)	0.094b
AVM	14,325	137 (0.96%)	2213	18 (0.81%)	0.52b
Dieulafoy lesions	14,325	6 (0.04%)	2213	1 (0.05%)	0.94b
Esophagitis	14,325	1,252 (8.7%)	2213	89 (4.0%)	< 0.001b
EoE	14,325	166 (1.2%)	2213	19 (0.86%)	0.21b
Atrophic mucosa	14,325	96 (0.67%)	2213	20 (0.90%)	0.22b
Barrett’s esophagus	14,325	641 (4.5%)	2213	113 (5.1%)	0.19b
Celiac disease	14,325	36 (0.25%)	2213	1 (0.05%)	0.056b
Candidiasis	14,325	80 (0.56%)	2213	8 (0.36%)	0.24b
GAVE	14,325	35 (0.24%)	2213	9 (0.41%)	0.17b
Melanosis	14,325	0 (0.0%)	2213	1 (0.05%)	0.13c
Mass	14,325	71 (0.50%)	2213	8 (0.36%)	0.39b
Nodules	14,325	318 (2.2%)	2213	55 (2.5%)	0.43b
Papule	14,325	169 (1.2%)	2213	32 (1.4%)	0.29b
Other lesions	14,325	464 (3.2%)	2213	78 (3.5%)	0.48b
Hiatal hernia	14,325	33 (0.23%)	2213	0 (0.0%)	0.024b
Diverticulum	14,325	4 (0.03%)	2213	1 (0.05%)	0.51c
Ring	14,325	163 (1.1%)	2213	25 (1.1%)	0.97b
Stricture	14,325	202 (1.4%)	2213	41 (1.9%)	0.11b
Varices	14,325	63 (0.44%)	2213	12 (0.54%)	0.50b
Food	14,325	3 (0.02%)	2213	0 (0.0%)	0.99c
Foreign body	14,325	256 (1.8%)	2213	1 (0.05%)	0.44c
Post-surgical anatomy variation	14,325	256 (1.8%)	2213	48 (2.2%)	0.21b
Z line	14,325	603 (4.2%)	2213	82 (3.7%)	0.27b

Statistics presented as n (column %)

p values: ^aanalysis of variance, ^cPearson’s Chi-square test, ^dFisher’s exact test

EGD esophagogastroduodenoscopy, AVM arteriovenous malformations, EoE eosinophilic esophagitis, GAVE gastric antral vascular ectasia

Colonoscopy

The colonoscopy mean time 17.9 (UL) vs. 22.3 min (LU), p < 0.001. Diagnostic indications account for 65.3% of all colonoscopies, followed by those performed for screening and surveillance purposes. When comparing different procedural sequences—UL versus LU—the LU sequence results in longer overall procedure time (17.9 ± 9.4 min vs. 22.3 ± 11.7 min, p < 0.001), time to reach the cecum (8.0 ± 6.8 min vs. 9.5 ± 6.9, p < 0.001), and withdrawal time (10.1 ± 6.4 min vs. 12.6 ± 7.3 min, p < 0.001). Additionally, clip placement was less frequent in the UL sequence (3.0% vs. 2.2%, p = 0.036). Colonoscopy for IBD surveillance biopsy was less preferred in the UL sequence (1.3% vs. 8.4%, p < 0.001) (Table 5).

Table 5.

Colonoscopy characteristics

Factor	Upper then lower (n = 14,325)	Statistics (upper then lower)	Lower then upper (n = 2213)	Statistics (lower then upper)	p value
Indication	14,325		2213		< 0.001b
Screening	14,325	2868 (20.0%)	2213	438 (19.8%)
Surveillance	14,325	1931 (13.5%)	2213	406 (18.3%)
Diagnostic	14,325	9526 (66.5%)	2213	1369 (61.9%)
Procedure duration (min)	14,325	17.9 ± 9.4	2213	22.3 ± 11.7	< 0.001a
Minutes to reach extent	11,440	8.0 ± 6.8	2015	9.5 ± 6.9	< 0.001a
Withdrawal time (min)	13,015	10.1 ± 6.4	2100	12.6 ± 7.3	< 0.001a
Maneuvers
Biopsy	14,325	4335 (30.3%)	2213	606 (27.4%)	0.006b
IBD surveillance biopsy	4335	57 (1.3%)	606	51 (8.4%)	< 0.001b
Microscopic colitis biopsy	4335	1877 (43.3%)	606	237 (39.1%)	0.051b
Dilation	14,325	24 (0.17%)	2213	6 (0.27%)	0.29b
Clips	14,325	425 (3.0%)	2213	48 (2.2%)	0.036b
Thermal therapy	14,325	108 (0.75%)	2213	21 (0.95%)	0.33b
Findings
Inflammation	14,325	553 (3.9%)	2213	86 (3.9%)	0.95b
Erosions	14,325	47 (0.33%)	2213	6 (0.27%)	0.66b
Ulcer	14,325	196 (1.4%)	2213	34 (1.5%)	0.53b
AVM	14,325	92 (0.64%)	2213	19 (0.86%)	0.25b
Melanosis	14,325	39 (0.27%)	2213	3 (0.14%)	0.23b
Polyps	14,325	5384 (37.6%)	2213	843 (38.1%)	0.65b
Nodules	14,325	59 (0.41%)	2213	6 (0.27%)	0.32b
Lipoma	14,325	59 (0.41%)	2213	6 (0.27%)	0.32b
Other lesions	14,325	146 (1.0%)	2213	22 (0.99%)	0.91b
Diverticulum	14,325	95 (0.66%)	2213	10 (0.45%)	0.24b
Hemorrhoids	14,325	12 (0.08%)	2213	5 (0.23%)	0.052b
Stricture	14,325	38 (0.27%)	2213	7 (0.32%)	0.67b
Varices	14,325	1 (0.01%)	2213	0 (0.0%)	0.99c

Statistics presented as n (column %)

p values: ^aanalysis of variance, ^bPearson’s Chi-square test, ^cFisher’s exact test

IBD inflammatory bowel disease, AVM arteriovenous malformations

Adenoma Detection Rate

When comparing the UL and LU sequences, the presence of polyps was similar (37.6% vs. 38.1%, p = 0.65), as was the overall ADR. Additionally, there was no statistically significant difference in ADR across the subgroups between the LU and UL sequences (Supplementary Table 3).

Discussion

In our large BDE cohort, performing EGD first resulted in shorter procedure times and a reduced likelihood of requiring opioids and benzodiazepines. Additionally, there were no differences between the two sequences regarding procedure-related adverse events or diagnostic yield, including ADR.

Same-day EGD and colonoscopy, i.e., BDE, are commonly performed as bundled procedures, but the optimal sequence for the procedure is not well established. The presented study found that performing EGD first has multiple benefits, such as requiring shorter procedural time and lower necessity of anesthesia, with similar diagnostic results and comparable adverse effects. The shorter procedural time is our most notable and innovative finding. Theoretically, performing EGD first allows for increased air insufflation in the bowel, which could potentially delay or complicate the subsequent colonoscopy [12]. However, this effect is minimized with CO₂ insufflation, which results in less abdominal distension compared to air insufflation [13]. A recent RCT concluded that CO₂ was superior to air insufflation regarding pain score in the hours after BDE, but the dose of sedative administered, analgesic usage, ADR, and adverse events were not statistically different between the groups [12]. Comparatively, another RCT comparing CO₂ and air insufflation in BDE recommended performing EGD first with CO₂ insufflation, due to similar quality of the procedure and adverse effects while requiring a lower dose of sedatives [14]. Our findings indicate that the shorter duration in the UL sequence suggests the impact of colon distension is not significant enough to hinder the performance of colonoscopy after EGD. Another hypothesis is that performing EGD first significantly reduced the overall duration of BDE procedures because colonoscopy can begin immediately after EGD while the patient is already sedated. During EGD, we typically aim for deeper sedation to minimize the gag reflex when the endoscope is inserted into the throat, ensuring a more comfortable procedure. We hypothesize that this may facilitate a quicker start to colonoscopy, as it reduces the need for additional sedation at the beginning of the procedure. In contrast to previous studies [7–11, 15, 16] that did not find significant differences in procedure times, our study demonstrated notable differences in EGD time, colonoscopy time, and overall combined duration. The colonoscopy observations indicated that the UL sequence resulted in a shorter time to reach the cecum and reduced withdrawal time, contributing to a quicker overall procedure. We hypothesize that the colonoscopy took longer than the UL procedures during LU procedures, as the colonoscopy performed during LU was primarily diagnostic and required more time compared to screening procedures. Consequently, utilizing the UL sequence not only shortens the procedure duration but also minimizes the need for extended sedation, highlighting a correlation between sedation duration and procedure time. The larger sample size in our research may have provided sufficient statistical power to identify these differences between the UL and LU approaches. Notably, this reduced time was also observed when procedures were performed by GI fellows (trainees), further validating our findings across various subgroups and clinical indications for BDE, both with and without fellows involved. As noted by Hsieh et al. [17], our findings support the notion that performing EGD followed by colonoscopy is the optimal sequence for same-day sedated BDE, enhancing the efficiency of the endoscopy unit.

Our study found that anesthesia (including general, MAC, and moderate sedation) was more commonly used in the UL sequence, while opiates and benzodiazepines were more frequently utilized in the LU sequence. Although most studies, including two recent meta-analysis [18, 19], have aimed to determine differences in total sedation doses between the UL and LU sequences—with only minor variations noted in some—our focus was on distinguishing the types of sedation and the likelihood of specific medications being used. Notably, the use of benzodiazepines, opiates, and antihistamines were generally similar between the groups. This may be interpreted by the fact that at the time to start the colonoscopy, and if the patient is already well sedated for the EGD, then the endoscopist can usually start the colonoscopy without any additional sedation and vice versa. The clinical significance of this finding, including its impact on recovery time and other outcomes, remains uncertain. However, it may inspire new studies aimed at evaluating these outcomes.

The EGD findings in our study were generally similar between groups, except for a higher incidence of esophagitis and hiatal hernia in the UL sequence, which was associated with reduced procedure time. In contrast, specific EGD procedures—such as feeding tube placement, Bravo tests, and video capsule placements—were performed more frequently in the LU cohort. These variations likely reflect the preferences of the endoscopists conducting the BDE, a factor that can impact retrospective studies. These variations are likely influenced by the preferences of the endoscopists performing the BDE in the UL approach to diagnosing upper GI symptoms. This variability can impact retrospective studies. Additionally, procedural fatigue associated with the colonoscopy-first (LU approach) may contribute to a lower incidence of esophagitis and hiatal hernia observed in the LU approach group. These findings suggest that patients presenting with upper GI symptoms, or a history of upper GI pathologies may benefit from undergoing an initial EGD.

Counterintuitively, certain colonoscopy maneuvers, such as clip placement, were performed more frequently in the LU sequence. Given the minimal difference between groups (3.0% versus 2.2%) and the lack of a clear clinical explanation, this discrepancy may be attributable to an unidentified bias.

The current colonoscopy surveillance guidelines in IBD [20] do not specify a preference for BDE. However, our study reveals that endoscopists generally prefer the LU sequence, probably because they believe it’s beneficial to start with the most pertinent procedures for the IBD population. This preference likely stems from the belief that starting with the most relevant procedures for the IBD population is advantageous. Given that specialists in a large American tertiary center preferred this sequence in a non-randomized study, it is likely that this sequence should be adopted until future studies recommend otherwise. Importantly, we found no difference in diagnostic yield capacity recorded in the procedure note. The diagnostic yield was the same in both study groups, which was expected, as the sequence of performing each procedure should not change the frequency of findings or yield. The ADR initially proposed by the Multi-Society Task Force [21] in 2002, is a critical measure of colonoscopy quality. Our study found no significant difference in ADR between the two groups, aligning with international findings [8, 9, 11, 16] and confirming that this approach maintains diagnostic quality. Evidence suggests that techniques such as retroflexion and additional forward views, which enhance mucosal visualization in unidirectional colonoscopy, likely benefit bidirectional endoscopy as well [22, 23]. Given that ADR is inversely associated with the risk of interval colorectal cancer, advanced-stage interval cancer, and fatal interval cancer [24], our study’s findings—showing no difference between sequences and an ADR exceeding the clinical practice guideline recommendation of at least 25% for average-risk colonoscopy [25, 26]—underscore the efficacy of BDE approach.

Additionally, adverse events recorded in the procedure note were similar between UL and LU sequence. Theatrically the adverse events could change—as performing a colonoscopy first could potentially increase aspiration risk, but in our cohort, we did not find a difference in terms of immediate procedure-related adverse events frequency that were documented in the procedure note. A caveat to this is that most endoscopy-related adverse events are transient, and the procedure note may not be reliable for detailed documentation.

Several factors could influence the decision on which procedure to perform first. Numerous studies, primarily from Asia [6, 7], and they have compared various elements such as procedure duration, the amount of sedation utilized, the diagnostic yield, procedure adverse events, and the patient’s discomfort. Importantly, previous studies have randomized the sequences and found similar diagnostic yield and adverse events between the groups. However, sedation requirements and procedure duration have significantly differed between the studied arms, with preference for EGD to be performed first. Although not described in our study, the discomfort score is another important variable and many studies show patients preference to the UL approach, including one of the few studies in Western population done in the United Kingdom [27].

This study offers several key strengths. First, it features a significantly larger patient cohort compared to most prior BDE studies. Second, this is the first large-scale U.S. study on this topic, offering key insights into shorter procedure times and a reduced need for opioids and benzodiazepines. Additionally, no significant differences were observed between the two sequences in terms of procedure-related adverse events or diagnostic outcomes, including ADR. Third, it provides unique insights not previously examined, including data on subgroups like patients with IBD, the involvement of trainees (such as gastroenterology fellows) in procedures, and the frequency of specific maneuvers and findings during EGD and colonoscopy at a U.S. tertiary care center. However, this study also presents some limitations, which would have been overcome if a RCT had been conducted: the observational nature of the design may have led to selection bias when comparing the various types of procedures, including provider preferences, before the procedure was done. We could not account for the colon preparation, as it is reported using different scales by various providers, and there was no reliable method to track this in a retrospective study. Another limitation is that our institution uses carbon dioxide (CO₂) for nearly all procedures, unless the unit is out of CO₂ or the procedures are conducted in the intensive care unit or operating room. Since this information is not documented, we are unable to account for it. However, we believe this effect was minimal, since both groups have similar demographics and procedures indications.

In our large BDE cohort, performing EGD first led to shorter procedure time and reduced need for opioids and benzodiazepines. There was no difference between study groups in terms of procedure-related adverse events and diagnostic yield, including ADR and overall outcomes. This study found that EGD first, followed by colonoscopy was more beneficial, which poses the question if standardization of EGD then colonoscopy may be more cost-effective, higher quality, and more time efficient compared to colonoscopy followed by EGD. Our study correlates with other studies performed internationally outside the US. However, further validation studies with prospective studies of our findings are needed, including looking at cost-effectiveness and transient procedure-related adverse event comparison between the two procedure sequences.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Jacqueline Estevez, and C. Roberto Simons-Linares (study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, critical revision), Arjun Chatterjee, Zurabi Zaalishvili, Renan Prado (drafting of the manuscript, critical revision), Rocio Lopez (Statistical analysis); John J. Vargo, Prabhleen Chahal, and Jean-Paul Achkar (Interpretation of data, critical revision).

Data Availability

No datasets were generated or analyzed during the current study.

Declarations

Conflict of interest

None of the authors have any relevant conflicts of interest.