To what extent can bowel preparation quality be considered “adequate”?: a systematic review and meta-analysis Authors: Chen-Ya Kuo, David Karsenti, Salvador Machlab, Fu-Jen Lee, Yu-Tsung Chen, Te-Ling Ma, Chi-Yang Chang, Yu-Tse Chiu Published: 2026 ABSTRACT AbstractBackground/AimsWhile low-quality bowel preparation is known to reduce the adenoma detection rate (ADR), the distinction between intermediate- and high-quality preparations remains unclear.MethodsElectronic searches were conducted in PubMed and Embase through July 2024. Randomized controlled trials reporting ADR and the Boston bowel preparation scale (BBPS) were included, and outcomes were pooled according to bowel preparation quality using a random-effects model. The primary outcome was ADR in intermediate-quality (BBPS ≥6, and ≥2 in each segment) versus high-quality (BBPS≥8) bowel preparation. Secondary outcomes included advanced ADR.ResultsFourteen trials were included (5,246 in the high-quality group and 1,728 in the other group). There was no significant difference in ADR (risk difference [RD], –0.02; p=0.13; I2=21%). Subgroup analyses showed no significant difference with computer-aided detection (CADe; RD, –0.03; p=0.70; I2=81%) or linked color imaging (RD, 0.04; p=0.31; I2=0%). The heterogeneity observed with CADe may reflect differences in colonoscopist experience or CADe systems. The results were similar for advanced ADR. Funnel plot and Egger’s test indicated minimal publication bias.ConclusionsIntermediate-quality preparation allows adequate adenoma detection. When each BBPS subscore is ≥2, surveillance recommendations can generally be followed.Keywords: Artificial intelligence, Cathartics, Colonic polyps, Colonoscopy, Image enhancement Graphical abstract INTRODUCTION Colorectal cancer (CRC) is a leading cause of death worldwide.1 Colonoscopy plays an important role in CRC screening programs and has been proven to decrease the incidence and mortality of CRC.2 Adequate bowel preparation improves the detection of colorectal lesions and is essential for successful screening colonoscopy.3 Current guidelines recommend an early repeat colonoscopy when bowel preparation quality is inadequate.4 While studies have shown a significantly lower adenoma detection rate (ADR) with low-quality preparation,5-8 the evidence does not indicate that high-quality preparation outperforms intermediate-quality preparation.9 Intermediate-quality bowel preparation may reduce mucosal visibility, potentially obscuring subtle lesions such as sessile serrated lesions.10 Conversely, the need for more extensive washing and suctioning may prolong withdrawal time and paradoxically increase the likelihood of adenoma detection through more careful inspection.11 As a result, the optimal level of bowel preparation quality remains uncertain. In addition, earlier studies predominantly used the Aronchick scale for bowel preparation, whereas the Boston bowel preparation scale (BBPS) has become more widely adopted in recent years. Scored from 0 (unprepared) to 1 (poor visibility), 2 (minor amount of residual stool but generally well visualized), and 3 (no residual fecal material) after flushing and suctioning, the advantage of the BBPS lies in its division of the colon into the right colon, mid-colon, and rectosigmoid colon, allowing for a more precise assessment of preparation quality in each segment.12 In addition, technological advances such as computer-aided detection (CADe) in colonoscopy and the linked color imaging (LCI) mode of Fujifilm endoscopy have both been proven to improve ADR,13,14 whereas the influence of different bowel preparation qualities on colonoscopy using these new technologies remains unknown. Given that the CADe model was developed using colonoscopist-annotated typical polyps15 and that stool, which appears bright yellow under LCI mode, can severely obstruct the colonoscopic view, it is worthwhile to investigate whether ADR differs between intermediate- and high-quality bowel preparation in CADe or LCI. Therefore, in this systematic review and meta-analysis, we aimed to investigate whether quality indicators (ADR/advanced ADR) differ between high-quality (BBPS 8–9) and intermediate-quality (BBPS 6–7 and ≥2 in each segment) bowel preparations. METHODS General guidelines We followed the latest Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines to conduct this systematic review and meta-analysis.16 This study was registered in PROSPERO under registration number CRD42024543818 and did not require ethics review board approval or informed consent from participants. Database searches and the identification of eligible manuscripts Two reviewers (Y.T.Chiu and C.Y.K.) independently conducted electronic searches in PubMed and Embase. The search was performed in July 2024 and covered studies published from January 2020 to July 2024. The search terms and detailed search strategy are provided in Supplementary Material 1. Inclusion and exclusion criteria The PICO (population, intervention/comparison, outcome) framework of this study was as follows: P, patients undergoing colonoscopy; I/C, high-quality and intermediate-quality bowel preparation (intermediate: BBPS≥6, and ≥2 in each segment; high: BBPS≥8); and O, ADR/advanced ADR. The inclusion criteria were randomized controlled trials (RCTs) enrolling human subjects undergoing colonoscopy that reported ADR and bowel preparation quality using the complete BBPS (meaning that either the subscores of all three segments or their total score had to be reported in the text). The exclusion criteria were: (1) non-RCTs; (2) non-English literature; (3) incomplete colonoscopy procedures (sigmoidoscopy was excluded); (4) no report of ADR (studies reporting only proximal or distal colon ADR were also excluded); (5) no report of complete BBPS; and (6) special study populations that might cause the results to differ from those of the general population, such as patients with hereditary polyposis syndrome. Disagreements regarding eligibility between the two reviewers were resolved through discussion until consensus was reached. Methodological quality appraisal Although the source studies were RCTs, the comparison of interest in this meta-analysis—bowel preparation quality—was not randomized, thereby disrupting the original randomization. In addition, ADR/advanced ADR stratified by BBPS were not the primary or secondary outcomes in most of these studies. The Cochrane risk-of-bias tool, which is outcome-specific and primarily designed to assess bias for prespecified trial outcomes, was therefore considered less suitable in this context. Accordingly, we adopted a modified Newcastle-Ottawa scale, an observational quality assessment tool, to assess the quality of each included study, as referenced in Toohey et al.17 Outcome The primary outcome was ADR stratified by intermediate-quality and high-quality preparation. The secondary outcome was advanced ADR stratified by bowel preparation quality, in which advanced adenoma was defined as an adenoma ≥1 cm, the presence of villous architecture, or high-grade dysplasia. Data extraction and management Data extracted from the study documents included the variables of interest (bowel preparation scale, ADR/advanced ADR), as well as the study protocol of each article for quality appraisal. One of the two reviewers (Y.T.Chiu) performed the data extraction, and the other (C.Y.K.) conducted the double-check. Disagreements were resolved through discussion. Owing to the numerical nature of the BBPS, it is often expressed as means/medians rather than as a categorical variable, which negatively affected data collection. Study investigators were contacted by e-mail for additional details when segment-level BBPS data were incomplete or not explicitly reported. Only studies with sufficient information to allow clear categorization of bowel preparation quality according to the prespecified definition (BBPS≥6 with a subscore ≥2 in each segment) were included. No assumptions were made when categorizing bowel preparation quality in the absence of adequate data. Statistical analysis RevMan ver. 5.4 (The Cochrane Collaboration) was used for statistical analysis. Because of the potential heterogeneity among the included studies, the meta-analysis was conducted using the inverse-variance method with a random-effects model (DerSimonian-Laird method).18 The effect size was measured using the risk difference (RD). Heterogeneity among studies was assessed using the I2 statistic; I2 values of 25%, 50%, and 75% were considered to indicate low, moderate, and high heterogeneity, respectively. To ensure the reliability of this meta-analysis, sensitivity analyses were conducted by systematically removing one study at a time. This approach was used to determine whether exclusion of any single study significantly altered the overall effect size. Prespecified subgroup analyses, defined a priori in the PROSPERO-registered protocol, were performed for colonoscopy arms using CADe or LCI. Potential publication bias was evaluated using a funnel plot generated by the software. Egger’s regression test was conducted to assess publication bias, using RD as the measure of effect size. The analysis was performed in Python using the statsmodels package. Trial sequential analysis (TSA) was performed using the TSA software package (available at http://www.ctu.dk/tsa/) to estimate the required sample size for the meta-analysis. RESULTS The PRISMA flowchart for the literature search is shown in Figure 1. Full-text articles excluded after eligibility assessment, along with the reasons for exclusion, are presented in Supplementary Table 1. Fourteen RCTs were included in the current analysis.19-32 Table 1 summarizes the participant composition, bowel preparation regimens, mean BBPS scores, and research protocol of each study. It should be noted that because patients with suboptimal bowel preparation quality (BBPS <6 or <2 in any segment) were included in some studies, the participant numbers listed here may differ slightly from those in the subsequent analyses. Appraisal using the modified Newcastle-Ottawa scale indicated generally good quality for most included studies (Table 1, Supplementary Table 2).19-32 Some limitations arose from the representativeness of the cases, as the participants (patients who failed bowel preparation during the index colonoscopy) or the colonoscopists (junior practitioners or procedures performed exclusively by the same colonoscopist) may not reflect general conditions.21,26,31 Furthermore, interventions in the original studies, such as CADe or LCI, could introduce confounding factors when comparing the intermediate- and high-quality groups in our meta-analysis.20,24,26 Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowchart. RCT, randomized controlled trial; ADR, adenoma detection rate; BBPS, Boston bowel preparation scale. Table 1.Study characteristicsStudyPublication yearLocationMale (%)Mean age (yr)Participants no.Preparation regimen/mean BBPSStudy design (colonoscopist/participant/equipment/intervention)Primary endpointsQuality appraisala)Machlab et al.192021Spain56.959.1836PEG+Asc(1+1)/NAExperienced/participants in the early CRC detection program/NA/single- vs. three-day low-residue dietBowel preparation quality8/8Hasegawa et al.202021Japan62.966.1700Sodium picosulfate 10 mL+1–2 L PEG/8.4Experienced/NP/Fujifilm EC-L600ZP(7)/tandem CS, the first time LCI vs. WLI followed by WLI in bothADR7/8Chen et al.212021China50.651.6346PEG(2+1)/7.3NAb)/NP/NA/with 6-minute VR video for patient education on bowel preparation or notBowel preparation quality7/8Sey et al.312022Canada55.160.7196PEG(2+2 or 4+2)+bisacodyl/NANA/patients who failed bowel preparation in index colonoscopy/NA/different preparation regimensBowel preparation quality7/8Sirinawasatien et al.222022Thailand60.058.7140PEG(1+1)+LB vs. PEG(2+2)/7.4Experienced/ NP/Olympus CF 180/different preparation regimensBowel preparation quality8/8Dao et al.232023Viet Nam49.142.0515PEG 3 L/7.2Experienced/NP/NA/mobile app for patient education on bowel preparation vs. controlBowel preparation quality8/8Karsenti et al.242023France48.658.42015NA/NAExperienced/NP/Olympus CF-H190, Fujifilm EC-760 R/V M/with or without CADeADR7/8Park et al.252023South Korea60.856.8171OST vs. 1 L PEG-A/7.9NA/NP/NA/different preparation regimensBowel preparation quality8/8Machlab et al.302025Spain51.558.9553PEG+Asc(1+1)/NAExperienced/participants of the CRC screening program/NA/unrestricted vs. 1-day low-residue dietBowel preparation quality8/8Lau et al.262024Hong Kong54.365.7766NA/7.85Junior/NP/Olympus CF-HQ290L, I series or CF-EZ1500DL, I series/with ENDO AID or notADR6/8Pattarapuntakul et al.282024Thailand46.256.7119PEG(2+2)/8.3Experienced/NP/Olympus CF-H190L/I or PCF-H190DL/mobile app for patient education on bowel preparation vs. controlBowel preparation quality8/8Abou Zeid et al.292024Lebanon54.654.0240PEG(3+1) vs. PEG(2+2)/7.6NA/NP/NA/different preparation regimensSide effects, tolerability, and willingness to retake the same preparation8/8Machlab et al.272024Spain59.959.3102PEG+Asc(1+1)/NANA/participants of the colorectal cancer screening program/NA/unrestricted vs. 3-day low-residue dietBowel preparation quality8/8Karsenti et al.322024France55.559.6686NA/NAExperienced/NP/Fujifilm high definition ELUXEO 700 series/LCI first vs. white light first in proximal colonAdenoma miss rate8/8Endoscopist experience level was categorized as experienced, junior, or not available (NA), based on information reported in the original articles in the “Study design” column, and was considered a potential source of heterogeneity.PEG, polyethylene glycol; Asc, ascorbic acid; NA, not available in the original article; CRC, colorectal cancer; NP, nothing particular (meaning that the study recruited participants with any kind of indication); CS, colonoscopy; LCI, linked color imaging; WLI, white light imaging; ADR, adenoma detection rate; VR, virtual reality; LB, lubiprostone; CADe, computer-aided detection; PEG-A, polyethylene glycol+ascorbate; OST, oral sodium sulfate tablet.a)Study quality appraisal by modified Newcastle-Ottawa scale, with the full marks of 8. b)All colonoscopies in this study were performed by a single colonoscopist; the level of experience was not specified in the article. Study characteristics Endoscopist experience level was categorized as experienced, junior, or not available (NA), based on information reported in the original articles in the “Study design” column, and was considered a potential source of heterogeneity. PEG, polyethylene glycol; Asc, ascorbic acid; NA, not available in the original article; CRC, colorectal cancer; NP, nothing particular (meaning that the study recruited participants with any kind of indication); CS, colonoscopy; LCI, linked color imaging; WLI, white light imaging; ADR, adenoma detection rate; VR, virtual reality; LB, lubiprostone; CADe, computer-aided detection; PEG-A, polyethylene glycol+ascorbate; OST, oral sodium sulfate tablet. a)Study quality appraisal by modified Newcastle-Ottawa scale, with the full marks of 8. b)All colonoscopies in this study were performed by a single colonoscopist; the level of experience was not specified in the article. In the pooled analysis of the 14 trials (Fig. 2A), there was no significant difference in ADR between intermediate- and high-quality bowel preparation, with low heterogeneity among studies (RD, –0.02; 95% confidence interval [CI], –0.05 to 0.01; p=0.13; I2=21%). In the sensitivity analysis using the one-study removal method, the results remained stable after exclusion of any individual study (Supplementary Fig. 1). Sensitivity analysis excluding the study with a lower quality score (6/8) yielded results consistent with the primary analysis (RD, –0.03; 95% CI, –0.06 to 0.00; p=0.08; I2=21%). Two included studies implemented CADe in their intervention arms,24,26 and two used LCI mode.20,32 To investigate the effect of bowel preparation quality on ADR under CADe/LCI, a subgroup analysis including these four trials was performed. ADR was significantly higher in the intermediate-quality group within the standard colonoscopy subgroup (RD, –0.03; 95% CI, –0.06 to –0.00; p<0.05; I2=0%), whereas no significant differences were observed in the CADe (RD, –0.03; 95% CI, –0.19 to 0.13; p=0.70; I2=81%) or LCI (RD, 0.04; 95% CI, –0.04 to 0.12; p=0.31; I2=0%) subgroups, despite substantial heterogeneity in the CADe subgroup (Fig. 2B). The funnel plot appeared roughly symmetrical around the dashed vertical line, suggesting no strong evidence of publication bias (Fig. 3). We also assessed publication bias using Egger’s regression test (Supplementary Fig. 2). The intercept was not statistically significant (intercept, –0.25; p=0.74), indicating minimal evidence of publication bias in our meta-analysis. TSA was conducted to assess whether the total sample size was sufficient to draw a definitive conclusion (Fig. 4). The cumulative Z-curve crossed the futility boundary and remained within the range of ±1.96, suggesting that further studies may not significantly alter the conclusion. (A) Forest plot of the effects of bowel preparation quality on adenoma detection rate (ADR); (B) Subgroup analysis of the computer-aided detection (CADe) arm, linked color imaging (LCI) arm, and standard colonoscopy. IV, inverse variance; CI, confidence interval. Funnel plot. SE, standard error; RD, risk difference. Trial sequential analysis. Six studies reported advanced ADR. The overall analysis showed no significant difference (RD, –0.02; 95% CI, –0.05 to 0.00; p=0.06; I2=16%) (Fig. 5A). Subgroup analysis indicated a higher advanced ADR in the intermediate-quality group within the standard colonoscopy subgroup (RD, –0.04; 95% CI, –0.08 to –0.01; p=0.02; I2=0%), but no significant differences were observed in the CADe (RD, 0.01; 95% CI, –0.03 to 0.05; p=0.57; I2=0%) or LCI (RD, –0.03; 95% CI, –0.10 to 0.03; p=0.34; I2=31%) subgroups (Fig. 5B). (A) Forest plot of the effects of bowel preparation quality on advanced adenoma detection rate (advanced ADR). (B) Subgroup analysis of the computer-aided detection (CADe) arm, linked color imaging (LCI) arm, and standard colonoscopy. IV, inverse variance; CI, confidence interval. DISCUSSION Adequate bowel preparation is one of the most important quality indicators of colonoscopy in the modern era.33 Atkin W et al.34 reported that poor preparation is associated with an increased CRC incidence rate, with a standardized incidence ratio of 1.30. In a post hoc analysis of the group with a BBPS score <2 in any segment, a very early repeat colonoscopy (median interval, 28 days) achieved a high mean ADR of 45.3% and an advanced ADR of 10.9%.35 These findings underscore the importance of adequate bowel preparation quality and the need for early repeat colonoscopy after inadequate bowel preparation. However, there is no guideline regarding the extent to which bowel preparation quality can be considered “adequate.” Menees et al. reported that 75% of patients with fair bowel preparation under the Aronchick scale were assigned recommendations inconsistent with guidelines, implying that a large proportion of colonoscopists do not consider intermediate-quality preparation to be “adequate”.36 On the other hand, our data indicated that intermediate-quality bowel preparation is comparable to high-quality preparation in terms of ADR (RD, –0.02; p=0.13) and advanced ADR (RD, –0.02; p=0.06). The cumulative Z-curve in TSA crossed the futility boundary, reinforcing the robustness of our findings and indicating that additional data are unlikely to provide sufficient power to demonstrate a clinically meaningful effect. Our findings, consistent with a large observational cohort,37 suggest that when the BBPS subscore is ≥2 in all segments, surveillance recommendations can generally be followed with less concern about missed polyps. Most included studies enrolled patients undergoing routine colonoscopy without restriction by indication, suggesting that the present findings are broadly applicable to both screening and non-screening populations. Nevertheless, caution may still be warranted in specific high-risk settings, such as patients undergoing colonoscopy for positive fecal tests or those with known hereditary syndromes, where further validation may be required. Notably, ADR and advanced ADR appeared higher in the intermediate-quality group, particularly in the standard colonoscopy subgroup in the subgroup analysis. The paradox of higher bowel preparation quality being associated with a lower ADR has been observed in previous studies.38-40 The authors attributed this to “overconfidence” in polyp detection under excellent bowel preparation. Aside from four studies that did not explicitly report withdrawal time,23,25,29,31 all other included studies reported median withdrawal times exceeding 6 minutes, in accordance with guideline recommendations.11 Sensitivity analysis excluding these four studies yielded similar results for overall ADR between intermediate- and high-quality bowel preparation (RD, −0.02; 95% CI, −0.05 to 0.01; p=0.15; I2=0%). Taken together, these findings suggest that inadequate withdrawal time is unlikely to have contributed substantially to the slightly lower ADR in the high-quality group. Still, it is important to highlight the additional benefits of high-quality preparation. For example, the literature indicates that high-quality preparation is associated with shorter insertion times, thereby improving procedural efficiency.41,42 Furthermore, a previous cost analysis demonstrated that imperfect preparation substantially increases colonoscopy costs.43 These advantages also align with key environmental, social, and governance (ESG) priorities. Therefore, the need for continuous quality improvement should be emphasized, and we should all keep in mind the importance of maintaining a careful examination with a withdrawal time exceeding 6 minutes regardless of bowel preparation quality.11 Although the number of included studies was limited, the findings indicate that the use of CADe or LCI did not substantially alter the outcomes. Notably, substantial heterogeneity was observed in the CADe subgroup analysis, which may be attributable to several study-level differences. The study by Lau et al.26 primarily involved junior colonoscopists, whereas Karsenti et al.24 enrolled experienced endoscopists. Previous studies have consistently shown that the incremental benefit of CADe is more pronounced among less experienced endoscopists,44,45 which may partly explain the divergent effect estimates between these two studies. In addition, different CADe systems were used (ENDO AID [Olympus] vs. GI Genius [Medtronic]), which may have further contributed to heterogeneity because of variations in algorithm design and performance. In contrast, the patient populations were broadly comparable, as both studies enrolled individuals undergoing routine colonoscopy. Taken together, these factors suggest that the findings of the CADe subgroup analysis are context-dependent and should be interpreted with consideration of endoscopist experience and CADe system characteristics. There are some limitations to our study. First, data collection relied on requests to corresponding authors, but the response rate was only about one-sixth. Consequently, nonresponse bias was inevitable. Second, although subgroup analysis of CADe/LCI-assisted colonoscopy was regarded as an important endpoint in this study, the number of included studies was limited because of the same issue mentioned in the first point. Finally, likely due to advances in bowel preparation methods, the high-quality preparation group substantially outnumbered the other group, potentially leading to reduced statistical power. Despite these limitations, our meta-analysis has notable strengths. First, we included trials from ten distinct areas, encompassing a variety of study aims and designs. Although these trials exhibited considerable diversity, heterogeneity in ADR across studies remained low, which increases the reliability of our pooled estimate. Second, all included trials were RCTs, ensuring high study quality, as reflected in their evaluation using the modified Newcastle-Ottawa scale. Although adenomas per colonoscopy (APC) is increasingly recognized as a standard quality indicator,46 it was not selected as our endpoint because of the limited availability of studies reporting APC. Moreover, as technological advances such as CADe, LCI, and distal attachment device-assisted colonoscopy continue to gain prominence, more data are needed to evaluate their performance in real-world settings and their interaction with bowel preparation quality. In conclusion, our meta-analysis demonstrated that intermediate-quality bowel preparation allows adequate adenoma detection. Similar findings were observed in CADe- and LCI-assisted colonoscopy, although these subgroup results should be interpreted in the context of study-level heterogeneity. Surveillance recommendations can generally be followed with less concern about missed polyps when the BBPS subscore is ≥ 2 in all segments. Importantly, these results should not be interpreted as diminishing the importance of meticulous mucosal inspection; rather, they underscore the need for colonoscopists to actively perform washing and suction during withdrawal when bowel preparation is not perfect but remains adequate. Future investigations incorporating APC or exploring emerging technologies, including CADe, LCI, and distal attachments, may provide additional insights into the impact of bowel preparation quality on colonoscopy outcomes. Footnotes Ethical Statements Not applicable. Conflicts of Interest The authors have no potential conflicts of interest. Funding None.