Risk factors for inadequate bowel preparation before colonoscopy: A meta‐analysis

This meta‐analysis aimed to comprehensively explore the risk factors for inadequate bowel preparation (IBP).


INTRODUCTION
Colonoscopy is the most effective diagnostic and screening method for various colorectal diseases.2][3] Successful colonoscopy depends largely on the quality of bowel preparation (BP).5][6][7] In addition, IBP can increase health care costs. 8Nonetheless, there is still a certain proportion of IBP in clinical practice. 9,10ndhi et al. and Mahmood et al. reviewed the literature up to 2016 and conducted meta-analyses on factors affecting IBP. 11,12wever, due to a lack of evidence, previous reviews included only factors related to patient characteristics rather than preparation-related factors and colonoscopy indications.Furthermore, the certainty of evidence was not reported in these studies.Hence, the current study aimed to comprehensively evaluate the risk factors for IBP and evaluate the certainty of evidence using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) method.

METHODS
This study was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. 13The protocol was registered in the International Prospective Register of Systematic Reviews database (CRD42022366498).

Search strategy
We searched the Embase, PubMed, Web of Science, and The Cochrane Library databases to identify studies examining risk factors for IBP.The search strategies are shown in the Supplemental Material.Additionally, we manually searched the bibliographies of eligible reviews and metaanalyses.The search was conducted on August 24, 2023.

Inclusion and exclusion criteria
The inclusion criteria were as follows: (1) observational studies or randomized controlled trials (RCTs) reporting at least one risk factor related to BP quality and (2) adult participants.Guidelines, reviews, meta-analyses, conference abstracts, case reports, editorials, and letters were excluded from the study.When more than one publication involved the same population, the most comprehensive report was included.We also excluded non-English-language papers for the pur-pose of referring to literature which was disseminated and read internationally.

Data extraction and risk of bias assessment
For each included study, the following data were extracted: first author, publication year, country, study design, number of sites, sample size, percentage of males, mean age and body mass index (BMI), percentage of white people, proportion of outpatients and screening/surveillance, frequency of IBP, BP regimen, dietary requirements, and quality scale.
In addition, we extracted the adjusted odds ratio (OR) and 95% confidence interval (CI) for each potential risk factor.The methodological quality was assessed using the Newcastle-Ottawa Scale (NOS) for observational studies and the Cochrane risk-of-bias 2 tool for RCTs.
Two investigators independently selected the studies, extracted the data, and assessed the risk of bias.Any discrepancies were resolved through discussion.

Statistical analysis
Due to different study designs, we conducted separate meta-analyses for observational studies and RCTs.If at least two studies reported the same exposure of interest, a random effects model was used to calculate the combined OR and 95% CI. 14 Heterogeneity across studies was detected using I 2 tests. 14Furthermore, subgroup analysis was performed based on the location of the study (Asia vs. non-Asia).If ten or more studies reported an outcome, the research characteristics presented in the previous section were used independently as explanatory variables for meta-regression analysis.On the premise of statistical significance, we stratified the data based on the elements accounting for the largest amount of heterogeneity and calculated the subset results separately.A funnel plot and Egger's test were used to explore publication bias, and statistically significant bias was adjusted by the trim-and-fill method. 15,16Additionally, sensitivity analyses were carried out by removing one study at a time.The significance level of all the statistical tests was p < 0.05.R 4.2.2 software was used for all the analyses.

Certainty of evidence
Each outcome was classified by magnitude: a relative risk of 0.75-1.25 was regarded as little-to-no association, 0.51-0.74and 1.26-1.99 were defined as small-to-moderate, and ≤0.50 or ≥2.00 were large.
Two reviewers used the GRADE approach to evaluate the credibility of Flow chart describing the study selection process.
evidence for each exposure-outcome correlation independently. 17e overall certainty of evidence was graded as high, moderate, low, or very low.According to the GRADE principle for prognostic studies, the evidence started at high 18,19 and was downgraded due to concerns about the risk of bias (studies with ≥50% weightage show a high risk of bias), indirectness (presence of factors that restrict the generalizability of findings), inconsistency across the studies (unexplained interstudy heterogeneity, I 2 ≥ 50%) or lack of consistency (single study), imprecision (95% CI indicating the effect may allow for multiple conclusions, for example, small-to-moderate and little-to-no association), and publication bias (substantial asymmetry observed in the funnel plot, p < 0.05 in Egger's test).

Study selection and characteristics
A total of 10,818 citations were initially screened, of which 502 full texts were reviewed.Finally, 125 articles (91 observational studies, 34 RCTs) were included in this meta-analysis (Figure 1).

Risk of bias assessment
The risk of bias assessment is shown in Tables S2 and S3 for observational studies and in Table S4 for RCTs.All observational studies were rated as moderate or high quality, with an average NOS score of 7.5.
Except for two RCTs that were assessed as high risk of bias, all other RCTs were assessed as low risk of bias or some concerns.

Risk factor analysis
The risk factors were combined and are presented in the following five parts: demographic variables, comorbidities, medications, preparation-related factors, and colonoscopy indications.The findings of observational studies and RCTs are summarized separately.

Demographic factors
The factors associated with a small-to-moderate increased in the risk of IBP included age (10-year increase; binary variable), BMI (binary variable), male sex, Medicaid insurance, and current smoking.However, age (continuous variable; 5-year increase), BMI (continuous variable), Medicare insurance, and past smoking had little-to-no association with IBP.Regarding ethnicity, our results revealed that, compared to white individuals, Asian ethnicity was associated with a small-to-moderate decrease in the risk of IBP, while Black ethnicity was associated with a small-to-moderate increase in the risk of IBP.Additionally, drinking, speaking English, and living alone were found to have little-to-no association with the risk of IBP.The evidence certainty of most factors was downgraded, mainly due to concerns related to inconsistency and imprecision (Figure 2).

Comorbidities
High certainty evidence indicated that Parkinson's disease, previous IBP, and poor mobility were associated with a small-to-moderate increase in the risk of IBP.In addition, a small-to-moderate increase in the risk of IBP was also detected for individuals with diabetes, liver cirrhosis, psychiatric disease (low certainty), inpatient, and Bristol stool form 1/2 (moderate certainty).The risk of IBP was not found to be associated with hypertension, American Society of Anesthesiology (ASA) ≥2/3 (low certainty), coronary artery disease (CAD), previous colonoscopy, or heart failure (moderate certainty) (Figure 3).

F I G U R E 3
Forest plot of the associations between comorbidities/medications and inadequate bowel preparation in observational studies.OR, odds ratio; CI, confidence interval; ASA, American Society of Anesthesiology.Explanations for GRADE: The evidence started at high certainty and was downgraded for (a) risk of bias, (b) inconsistency/lack of consistency, (c) indirectness, (d) imprecision, and (e) publication bias.Lowercase and capital letters indicate a decrease of one or two levels, respectively.No conclusions were reported regarding results with very low certainty.The conclusions were based on the magnitude of the effect: little-to-no association (gray), small-to-moderate association (orange), and large association (red).No conclusions were reported regarding results with very low certainty (white).

Medications
We had high certainty that antipsychotic use was a small-to-moderate risk factor for IBP.Additionally, our results indicated that other drugs associated with a small-to-moderate increase in the risk of IBP include tricyclic antidepressants, narcotics, calcium channel blockers (CCBs) (moderate certainty), opioids, and antidepressants (low certainty).
Low certainty evidence suggested that benzodiazepines, laxatives, and the number of medications had no relationship with IBP risk (Figure 3).

Preparation-related factors
High certainty evidence indicated that characteristics of last stool (solid or brown liquid) were a strong risk factor.Furthermore, incomplete preparation intake (moderate certainty) and incorrect diet restriction (low certainty) were also strong risk factors.Small-tomoderate risk factors included preparation-to-colonoscopy interval not within 3 to 5/6 h (high certainty), preparation-to-colonoscopy interval (binary variables), nonsplit preparation, and preparation instructions not followed (moderate certainty).Low certainty evidence suggested that afternoon colonoscopy, fellow involvement, and preparation-to-defecation interval were unrelated to the risk of IBP (Figure 4).

Indications for colonoscopy
Low certainty evidence showed that screening/surveillance colonoscopy, colonic polyps, positive fecal occult blood test, and bleeding were not associated with the risk of IBP.There was very low certainty for several other indications (Figure 4).

RCTs
Some findings from observational studies had very low certainty but showed higher certainty in RCTs, including education (small-tomoderate, low certainty), constipation (small-to-moderate, moderate certainty), stroke/dementia (small-to-moderate, low certainty), previous surgery (little-to-no association, low certainty), preparationto-colonoscopy interval (per hour increase) (small-to-moderate, low certainty), and discomfort during preparation (small-to-moderate, high certainty).
In contrast to observational outcomes, four RCTs reported that BMI (binary variable) was not related to the risk of IBP (low certainty).
Furthermore, for inflammatory bowel disease (IBD), ASA score, and inpatient, the results from RCTs were also different from those from observational studies.However, each had only one RCT assessment, making it impossible to conduct meta-analyses.Other findings were in line with the meta-analyses of observational studies (Figure 5).

Subgroup analysis, meta-regression analysis and sensitivity analysis
Tables S5 and S6 summarized the subgroup analyses of location.
For factors that were reported in ten or more studies, we further F I G U R E 5 Forest plot of the associations between factors and inadequate bowel preparation in randomized controlled trials.OR, odds ratio; CI, confidence interval; ASA, American Society of Anesthesiology.*The cutoff points of age included 50, 60, and 65 years old.† The cutoff points of body mass index included 26 and 30 kg/m 2 .‡ The cutoff points of preparation-to-colonoscopy interval included 5, 6, and 7 h.Explanations for GRADE: The evidence started at high certainty and was downgraded for (a) risk of bias, (b) inconsistency/lack of consistency, (c) indirectness, (d) imprecision, and (e) publication bias.Lowercase and capital letters indicate a decrease of one or two levels, respectively.No conclusions were reported regarding results with very low certainty.The conclusions were based on the magnitude of the effect: little-to-no association (gray), small-to-moderate association (orange), and large association (red).No conclusions were reported regarding results with very low certainty (white).
conducted subgroup analyses based on the meta-regression results.
Notably, in observational studies, the heterogeneity of sex, hypertension, inpatient, and afternoon colonoscopy could not be elucidated by any of the variables in the meta-regression.For sex, constipation, and diabetes in RCTs, we did not conduct additional subgroup analyses because no heterogeneity was detected among the studies.The details can be found in Tables S5 and S6, and the corresponding forest plots are shown in Figures S1-S23.
As listed in Tables S7-S65, leave-one-out meta-analyses showed that most pooled effects did not change substantially.However, in observational studies, for drinking, psychiatric disease, CAD, ASA score ≥2/3, previous colonoscopy, previous IBP, laxatives, number of medications, preparation-to-defecation interval, IBD, and bleeding, the associations between each of these findings and the risk of IBP might have been driven by a single study.Similar results were also found for education and incorrect diet restriction in RCTs.Moreover, for CCBs and preparation instructions not followed in observational studies, the absence of associations between the above two factors and the risk of IBP was caused by one study.The same applied to age in RCTs.

Publication bias
In the meta-analyses of observational studies, risk factors, including BMI, insurance, current smoking, drinking, constipation, diabetes, stroke/dementia, previous surgery, liver cirrhosis, tricyclic antidepressants, incorrect diet restriction, and preparation-to-colonoscopy interval, showed evidence for publication bias according to the funnel plots (Figures S24-S31) and Egger's test (Table S5).In the metaanalyses of RCTs, publication bias was also found in preparationto-colonoscopy interval (Table S6, Figures S32 and S33).After the trim-and-fill approach was used, the statistical significance of current smoking, drinking, diabetes, previous surgery, and incorrect diet restriction did not change, however, the meanings did change for other factors (Figures S34-S46).Previous meta-analyses had shown that increased age was a risk factor for IBP; however, the results regarding BMI had been contradictory. 11,12In addition, neither of them conducted subgroup analyses based on different units.Our meta-analyses of observational studies indicated that age (binary variable) and BMI (binary variable)

DISCUSSION
were moderate risk factors, while age (continuous variable) and BMI (continuous variable) were not.Notably, pooled analyses of the RCTs revealed no association between BMI (binary variable) and IBP risk, and further prospective research was needed.In addition, a moderate increased risk was observed for males and current smokers.One study noted that male patients were less likely to follow preparation instructions than female patients 20 ; however, whether this could explain the positive relationship was still unknown.Medicaid insurance recipients and lower education status were also moderate predictors of IBP.One possible assumption of this association was that noncompliance might differ among different socioeconomic groups.For race, compared with whites, Asians had lower rates of IBP, and Black people had higher rates, which could be due to differences in BMI, pharmacotherapy, eating habits, and access to health care.
Constipation and Bristol stool type 1/2, indications of slow colonic transit or defecatory disorders, appeared to be moderately associated with IBP.2][23] Notably, one RCT reported inconsistent results with observational studies for inpatient 24 and ASA score. 25Neurological conditions (Parkinson's and psychiatric disease) were also associated with IBP.Patients with diminished mental capacity might need to help understanding and following preparation instructions.Furthermore, previous IBP was also a risk factor; however, previous colonoscopy was not.
7][28][29][30] Similar to previous studies, our analyses for opioid and narcotic use revealed a moderate increased risk of IBP.
The results of earlier reviews for CCBs were contradictory, 11,12 and our findings were in line with Gandhi's study.Tricyclics and other antidepressants were also positively correlated with IBP.Depression itself had been proven to be related to gut dysfunction, 31 and antidepressant drugs affect gastrointestinal motility through anticholinergic effects. 32,33Taking antipsychotics increased the occurrence of IBP.
On the one hand, it might not be easy for psychiatric patients to follow preparation instructions.On the other hand, constipation was a common adverse effect of all antipsychotic drugs. 34,35r observational meta-analyses indicated that the three strongest risk factors for IBP were characteristics of last stool (solid or brown liquid), incomplete preparation intake, and incorrect diet restriction, all of which were related to the colonoscopy preparation process.
In addition, the split-dose regimen had been recommended by many studies. 36,37Our study provided further evidence of the advantage of a split-dose regimen.Bucci et al. suggested that the superiority of the split dose was prominently related to the preparation-to-colonoscopy interval. 37In the present analysis, waiting time of 3-5/6 h was beneficial.Finally, few studies evaluated the effects of various colonoscopy indications, and our results showed that none of them was associated with IBP.More exploration is needed in the future.
Suboptimal BP is a hodgepodge of many situations.The collective denominator for most factors seems to be decreased bowel motility or poor compliance.Our study systematically identified patient-related factors from four aspects: demographic variables, comorbidities, medications, and colonoscopy indications.It is possible to prevent colonoscopy failure by implementing enhanced preparation regimens for at-risk patients identified based on patient characteristics.9][40] Our findings might provide a reference for optimizing predictive models in the future.
In general, patient-related factors seemed to be moderately correlated with IBP.In contrast, preparation-related elements seemed to have stronger associations with IBP.This discovery underscores the importance of compliance with dietary modifications and oral laxatives.
A study reported that nearly 20% of patients with IBP did not follow preparation instructions. 41Medical staff should provide written and oral preparation instructions and emphasize the significance of compliance.In addition, various tools, such as cartoon pictures, telephone calls, educational videos, and social media applications, could be used to enhance instructions, which had been proven to increase patient compliance and improve BP quality. 42,43Finally, the characteristics of last stool (solid or brown liquid) were an intuitive response to IBP, and evaluating feces before the operation was necessary. 44Recently, two studies proposed smartphone apps to assist patients in judging BP status through the appearance of feces; however, there was still a gap in widespread clinical application. 45,46e current research had several advantages.First, this was an extensive review of the factors affecting IBP to date, and it was the first meta-analysis of factors related to colonoscopy indications and preparation processes.Second, our conclusions were based on the magnitude of the association reaching predefined thresholds.Third, we evaluated the quality of evidence using the GRADE approach.
Nevertheless, there were still some limitations.First, although we conducted subgroup analyses based on meta-regression and region, many factors still exhibited significant heterogeneity.Therefore, the random effects model was selected.Heterogeneity was expected.On the one hand, BP methods, including purgative volume and types, dietary restriction, scales, geographical locations, population characteristics, and study design, were quite different among the included articles.On the other hand, although adjusted ORs were used, the adjustment usually included distinct variables in each study.Second, publication bias existed.After performing the trim-and-fill method to adjust the results, the statistical significance of BMI, insurance, constipation, stroke/dementia, liver cirrhosis, tricyclic antidepressants, and preparation-to-colonoscopy interval changed.However, the results were robust after sensitivity analyses.Hence, these findings should be interpreted with caution.Third, the definitions of constipation and surgery were inconsistent across studies.Fourth, the current review only included studies written in English, which might introduce a language bias.Notably, although only studies published in English were considered, one-fifth of them (26 studies) were from China.We believed that the results could also represent the situation in China.
Finally, due to these limitations, the certainty of evidence was very low for some outcomes, making it difficult to draw valid conclusions.
The obvious risk factors for IBP included three preparation-related factors: characteristics of last stool (solid or brown liquid), incomplete preparation intake, and incorrect diet restriction.Demographic, disease, and medication factors were mainly moderately correlated with IBP.No colonoscopy indications are found to be related to IBP.These findings may help clinicians recognize high-risk individuals and provide guidance for IBP prevention and future research.
Forest plot of the associations between demographic variables and inadequate bowel preparation in observational studies.OR, odds ratio; CI, confidence interval.*The cutoff points of age included 40, 50, 54, 55, 60, 61, 65, and 75 years old.† The cutoff points of body mass index included 25, 27, and 30 kg/m 2 .Explanations for GRADE: The evidence started at high certainty and was downgraded for (a) risk of bias, (b) inconsistency/lack of consistency, (c) indirectness, (d) imprecision, and (e) publication bias.Lowercase and capital letters indicate a decrease of one or two levels, respectively.The conclusions were based on the magnitude of the effect: little-to-no association (gray), small-to-moderate association (orange), and large association (red).No conclusions were reported regarding results with very low certainty (white).

F I G U R E 4
Forest plot of the associations between preparation-related factors/colonoscopy indications and inadequate bowel preparation in observational studies.OR, odds ratio; CI, confidence interval; FOBT, fecal occult blood test.*The cutoff points of preparation-to-colonoscopy interval included 4, 5, 6, and 8 h.Explanations for GRADE: The evidence started at high certainty and was downgraded for (a) risk of bias, (b) inconsistency/lack of consistency, (c) indirectness, (d) imprecision, and (e) publication bias.Lowercase and capital letters indicate a decrease of one or two levels, respectively.No conclusions were reported regarding results with very low certainty.The conclusions were based on the magnitude of the effect: little-to-no association (gray), small-to-moderate association (orange), and large association (red).No conclusions were reported regarding results with very low certainty (white).