Abstract
Background
Colorectal cancer is a leading cause of cancer-related death. Adenomas and serrated polyps are precursors of colorectal cancer, with serrated polyps being more difficult to detect during colonoscopy. The relationship between propofol use and polyp detection remains unclear. We investigated the association of propofol-based versus mild-moderate sedation on adenoma and serrated polyp detection during colonoscopy.
Methods
This retrospective cohort study used observational data from the New Hampshire Colonoscopy Registry. Patients aged over 50 with screening or surveillance colonoscopies between 1/1/2015 and 2/28/2020 were included. Exclusions were diagnostic exams, no sedation, missing pathology data, and poor bowel preparation. Multivariate logistic regression was used to evaluate differences in polyp detection between propofol and moderate sedation in the full sample, while adjusting for covariates. Propensity score adjustment and clustering at the endoscopist level was used in a restricted sample analysis which included endoscopists and facilities with between 5% and 95% propofol sedation use.
Results
54,063 colonoscopies were analyzed in the full sample and 18,998 in the restricted sample. Serrated polyp prevalence was significantly higher using propofol (9957/29312; 34.0% [CI 33.4–34.5]) vs. moderate sedation (6066/24751; 24.5% [CI 24.0–25.1]) in the full sample and restricted samples (1410/4661; 30.3% [CI 28.9–31.6] vs 3690/14337; 25.7% [CI 25.0–26.5]). In the full sample multivariate logistic regression, propofol was associated with higher neoplasm (adjusted odds ratio [OR] 1.25 [1.21–1.29]), adenoma (OR 1.07 [1.03–1.11]), and serrated polyp detection (OR 1.51 [1.46–1.57]). In the restricted sample using inverse probability of treatment weighted propensity score adjustment and clustering at the endoscopist level, an attenuated but statistically significant effect size was observed for serrated polyps (OR 1.13 [1.07–1.19]), but not for adenomas (OR 1.00 [0.95–1.05]) or any neoplastic lesion (OR 1.03 [0.98–1.08]).
Conclusion
Propofol sedation during colonoscopy may be associated with improved detection of serrated polyps, but not adenomas.
Summary Statement:
Compared to moderate sedation using benzodiazepines and opioids, propofol-based sedation during colonoscopy may be associated with higher likelihood of detecting serrated polyps compared to moderate sedation. There was no difference in adenomatous polyp detection.
INTRODUCTION
Colorectal cancer is a global public health concern as the second leading cause of cancer-related death both in the United States and worldwide.1–3 Colonoscopy is a highly effective screening tool for colorectal cancer, allowing for the removal of adenomas and serrated polyps, the two major precursors. Endoscopic identification helps prevent their progression to malignancy. Adenomas are more common and tend to be easier to detect due to their size, raised appearance, and location in the colon.4,5 Serrated polyps are considered more difficult to identify, since they are typically located in the proximal colon and are often flat with indistinct borders blending into surrounding colonic tissue.6–8
Identifying measures to optimize the detection of adenomas and serrated polyps during colonoscopy is critical for colorectal cancer prevention. Only a handful of studies have sought to determine the relationship between propofol-based deep sedation and precancerous polyp detection during colonoscopy. Overall, the results of these studies have been conflicting and subject to significant limitations, including small sample sizes, missing adjustment for pertinent patient and procedural factors, and absence of pathology data for polyp classification.9–14 Given that serrated polyps pose greater challenges to endoscopic visualization than adenomas, enhanced performance conditions may theoretically improve the detection of serrated lesions compared to adenomas.
The purpose of our study was to use the population-based New Hampshire Colonoscopy Registry to investigate the relationship between propofol based sedation compared to mild-moderate conscious sedation on overall polyp detection during screening and surveillance colonoscopy. Furthermore, we sought to evaluate the association of propofol sedation and the detection of serrated lesions compared to detection of adenomas. Prior studies examining the association of propofol sedation with polyp detection did not differentiate by polyp histology, despite the challenges in identifying serrated lesions. We hypothesized that the improved conditions afforded by propofol use can enhance overall polyp detection, with most pronounced effects on the detection of serrated lesions which pose the greatest challenge for endoscopic visualization.
METHODS
Following Institutional Review Board approval (Committee for the Protection of Human Subjects at Dartmouth College # 15834), we performed a retrospective cohort study using data obtained from the New Hampshire Colonoscopy Registry prospective registry. The New Hampshire Colonoscopy Registry is a statewide, population-based colonoscopy registry which collects comprehensive colonoscopy data from patients, physicians, and pathology labs in New Hampshire. The New Hampshire Colonoscopy Registry database currently contains detailed data on patient risk factors, colonoscopy characteristics, endoscopists, and procedural quality indicators (which include bowel preparation quality, withdrawal time, exam completions, and cecal intubation) as well as polyp histology (e.g. serrated and adenomatous polyps) for nearly 275,000 colonoscopies in over 205,000 patients. All examinations in the New Hampshire Colonoscopy Registry were performed within 30 endoscopy practices in New Hampshire. See text document, Supplemental Digital Content 1, providing detailed information on the data collection process of the New Hampshire Colonoscopy Registry.
A data analysis and statistical plan was written and filed with the Dartmouth-Health Office of Research Operations and with Maine Medical Center before data were accessed. The STROBE checklist, designed for observational studies, was followed in the preparation of this manuscript.15
Our study included colonoscopies which were complete to the cecum with adequate bowel preparation in patients aged 50 and older with a colonoscopy performed between January 1st, 2015 and February 1st, 2020. This study period was prior to a recent change by the United States Preventive Services Task Force recommending screening colonoscopy to begin at age 45.16 Our sample included both screening exams, in which patients have no prior history of precancerous colon polyps or colorectal cancer, and surveillance exams, in which patients do have a prior history of polyps or colorectal cancer. We included both screening and surveillance exams in our cohort as the type of sedation used would have similar effect on screening and surveillance exams, independent of indication. We excluded exams with incomplete data on key exam characteristics: exam indication, type of sedation, exams with missing pathology reports, and exams that were not complete to the cecum (Figure 1). Surveillance colonoscopies for genetic syndromes or inflammatory bowel disease, all diagnostic colonoscopies (e.g., for anemia, bleeding, or diarrhea) and exams with poor bowel preparation or those performed without the use of sedation (uncommon in the U.S.) were excluded from our analysis. Poor (inadequate) bowel preparation is defined on each data collection form as “feces and/or nontransparent fluid, definitely impairing visualization”.17 Given the rarity of complete exams with poor bowel preparation in New Hampshire Colonoscopy Registry data, we assumed complete exams with missing preparation had adequate preparation. Otherwise, we performed multiple imputation using chained equations (MICE) to impute missing covariates (e.g. BMI, smoking status). Data was not imputed for outcome variables, as all exams with missing outcome data were excluded from the analysis (Figure 1).
Figure 1:
CONSORT Diagram
Patients were assigned to either the propofol-based sedation or moderate sedation only cohort based on the sedation method used during colonoscopy. Patients unable to tolerate a complete examination under moderate sedation are rescheduled for an examination using propofol with anesthesia personnel, and only complete exams were included in our analysis.
Outcomes, Treatment and Covariates:
Outcomes evaluated were (i) neoplastic polyps, defined as the presence of any adenomatous and /or serrated polyps, (ii) adenomatous polyps, and (iii) serrated polyps. Our primary exposure of interest was the use of propofol-based sedation versus moderate sedation using benzodiazepines and opioids.
Continuous age, continuous body mass index, patient sex, alcohol use, smoking status, self-reported health, family history of colorectal cancer in a first degree relative, patient history of previous colonoscopy, patient history of previous neoplastic polyp findings, and exam preparation quality were identified as covariates and potential confounders. We also considered the influence of facility and endoscopist. See table, Supplemental Digital Content 2, for a data dictionary of the variables included in our analysis.
Statistical Methods:
In our full sample analysis, after application of all sample restrictions using exams from all facilities and endoscopists regardless of how frequently they used propofol, we performed three sets of regression analyses. First, we employed univariate logistic regressions for (i) any neoplastic polyp detection, (ii) adenomatous polyps, and (iii) serrated polyps on propofol vs. moderate sedation use. We then used a multivariable logistic regression to adjust for possible confounding due to the following covariates: continuous age, continuous body mass index, patient sex, alcohol use, smoking status, self-reported health, family history of colorectal cancer in a first degree relative, patient history of previous colonoscopy, patient history of previous neoplastic polyp findings, and exam preparation quality. We also ran a generalized linear mixed effects model which adjusted for the same possible confounders and accounted for clustering of patients within endoscopy practice sites using random effects. In a separate set of models, we accounted for clustering at the endoscopist level within the full sample, but found model convergence to be sensitive to the choice of optimization algorithm. See table, Supplemental Digital Content 3, for an overview of the model specifications and covariates used in our full sample analysis.
Because 24 facilities [75%] and 80 endoscopists [53%] in our full sample used propofol almost never (<5%) or nearly exclusively (>95%), it was difficult to disentangle the relationship between propofol use and polyp outcomes from the role of other endoscopist or facility-level characteristics among those endoscopists and facilities. Since this limited our ability to identify the influence of propofol on increased polyp detection, we also conducted a restricted sample analysis of exams performed only at facilities and by endoscopists who varied their use of propofol. This restricted sample included only those endoscopists and facilities who used propofol for between 5% and 95% of their exams.
In this restricted-sample analysis, we performed six separate sets of regressions. Refer to table, Supplemental Digital Content 4, for a comprehensive overview of the model specifications and covariates used for the restricted sample. First, we ran univariate logistic regressions as in the full sample analysis, and then followed with two different sets of regression models described below, one using propensity scores, converted into a 20-ile and entered as a categorical variable, and one using inverse probability of treatment weighting, to reduce the potential effects of confounding in this observational study. Inverse probability of treatment weighting was used in addition to propensity categories to ensure that our findings were robust and not sensitive to how the propensity score was used. We then repeated each of these three sets of regressions using generalized linear mixed effect models with endoscopist-level random effects. We also performed a separate regression as a sensitivity analysis where we added time in years as a categorical covariate to the inverse probability of treatment weighted models Refer to table, Supplemental Digital Content 5, for an overview of the adjustments used in our study and detailed information regarding the rationale for implementation.
The purpose of this restricted sample analysis was to satisfy the positivity assumption of causal inference, by analyzing only endoscopists and facilities who had a non-zero probability of using propofol for some exams and moderate sedation for others.
Propensity scores are intended for situations where there is a chance that each event will receive either type of treatment; in this study, this includes those exams at facilities and by endoscopists who used both propofol and moderate sedation. For patients in the full sample who received care at facilities or from endoscopists where propofol was used either nearly exclusively (> 95%) or almost never (< 5%), this was not the case. We therefore did not include regressions using either propensity score categories as a covariate or inverse probability of treatment weighting in our full-sample analysis.
The inverse probability of treatment weighting balances both the subjects who received propofol and those who received moderate sedation to reflect the combined population. We calculated propensity scores by modeling the probability of propofol (vs moderate sedation) in terms of the covariates age, BMI, sex, alcohol use, smoking status, self-reported health, first degree relative with history of colorectal cancer, previous colonoscopy, history of previous neoplastic lesions, and bowel preparation quality using logistic regression. We graphically compared the distributions of the resulting propensity scores in the propofol and moderate sedation groups to ensure a large region of common support. The inverse probability of treatment weighted models assigned weights to each observation based on the inverse of the estimated propensity score. We also calculated standardized mean differences (i.e. difference in means divided by pooled standard deviation) for patient and exam characteristics before and after weighting the moderate sedation and propofol groups by the inverse of the probability of treatment (i.e. the propensity score) to make sure that we had successfully adjusted for the differences between the two groups. All analyses were performed in R (version 3.6) using the mice, lme4, tidyverse, and broom.mixed packages.
RESULTS
A total of 54,063 endoscopic examinations met our inclusion criteria for analysis. Of those, 54.22% were performed using propofol (29,312 exams). Table 1 describes the absolute, unweighted characteristics of the propofol and moderate sedation samples in both the full and the restricted sample analyses. Table 1 also reports standardized mean differences between samples before and after inverse probability of treatment weighting using propensity scores. Patient demographics (e.g. age, sex, BMI) were evenly distributed among patients who received propofol compared to moderate sedation, with all differences falling around the one-unit or 1 percentage point range. Specifically, the mean age (standard deviation) was 61.2 years (7.9) for the propofol group, compared to 62.3 years (8.2) for the moderate sedation group. Similarly, the mean BMI (SD) was 29.3 (7.5) for the propofol group and 28.3 (6.8) for the moderate sedation group. Clinical and procedural characteristics also demonstrated a balanced distribution between the two groups, with none exhibiting a difference of more than 5%. Notably, the largest difference observed was with bowel preparation quality, which was lower in the propofol group, with 33.1% of exams in the moderate sedation group having excellent bowel preparation, compared to 28.1% in the propofol group.
Table 1:
Patient Characteristics in both the Full and Restricted samples
| Full Sample N=54063 | Restricted Sample1 N=18998 | |||||
|---|---|---|---|---|---|---|
| Propofol N/% or mean (SD) | Moderate Sedation N/% or mean (SD) | Propofol N/% or mean (SD) | Moderate Sedation N/% or mean (SD) | Standardized mean differences2 | ||
| No inverse probability of treatment weighting | Inverse probability of treatment weighting | |||||
| Age | 61.2 (7.9) | 62.3 (8.2) | 63.0 (8.3) | 62.8 (8.4) | 0.022 | 0.017 |
| BMI 3 | 29.3 (7.5) | 28.3 (6.8) | 30.7 (8.7) | 27.9 (6.6) | 0.367 | 0.013 |
| Patient sex | ||||||
| Female | 14978 (51.1) | 12542 (50.7) | 2412 (51.8) | 7009 (48.9) | −0.647 | −0.007 |
| Family history of colorectal cancer in first degree relative 3 | ||||||
| Yes | 5840 (23.5) | 5309 (24.7) | 1124 (28.8) | 3075 (24.9) | −0.371 | 0.016 |
| Patient had previous exam | ||||||
| Yes | 19749 (67.4) | 17705 (71.5) | 3516 (75.4) | 10413 (72.6) | −0.813 | 0.019 |
| Patient had previous neoplastic findings | ||||||
| Yes | 11048 (37.7) | 9037 (36.5) | 2187 (46.9) | 5995 (41.8) | −0.503 | 0.009 |
| Regular4 aspirin use | ||||||
| Yes | 7553 (30.3) | 6890 (31.8) | 1451 (36.6) | 4094 (32.6) | −0.444 | 0.018 |
| Regular4 NSAID use | ||||||
| Yes | 3833 (15.0) | 3535 (16.0) | 683 (16.8) | 2042 (15.9) | −0.302 | −0.003 |
| Regular4 blood thinner use | ||||||
| Yes | 614 (2.3) | 503 (2.2) | 172 (4.0) | 303 (2.3) | −0.067 | 0.044 |
| Smoking status 3 | ||||||
| Current | 2268 (8.7) | 1371 (6.1) | 427 (10.3) | 778 (6.0) | −0.116 | 0.011 |
| Never | 13674 (52.5) | 12663 (56.3) | 1940 (46.8) | 7437 (56.9) | −0.768 | −0.018 |
| Alcohol use 3 | ||||||
| More than 8 drinks per week | 2665 (10.3) | 2712 (12.1) | 430 (10.4) | 1629 (12.5) | −0.299 | 0.009 |
| More than 20 drinks per week | 338 (1.3) | 259 (1.2) | 66 (1.6) | 155 (1.2) | −0.066 | 0.002 |
| General health 3 | ||||||
| Poor health | 146 (0.6) | 57 (0.3) | 65 (1.6) | 38 (0.3) | 0.026 | 0.011 |
| Good or better health | 24318 (93.4) | 21320 (95.3) | 3551 (85.2) | 12362 (94.9) | −1.186 | −0.016 |
| Surveillance indication | 10527 (35.9) | 8230 (33.3) | 2000 (42.9) | 5440 (37.9) | −0.469 | 0.012 |
| Insurance | ||||||
| Medicaid/Medicare | 8745 (31.1) | 8333 (34.7) | 2030 (45.4) | 5200 (37.6) | −0.448 | 0.087 |
| Private/Health maintenance organization | 18578 (66.0) | 16263 (67.8) | 2503 (55.9) | 9144 (66.1) | −0.845 | −0.123 |
| Other/Unsure | 2502 (8.9) | 2121 (8.4) | 512 (11.4) | 1298 (9.4) | −0.200 | 0.045 |
| No insurance | 75 (0.27) | 76 (0.32) | 15 (0.34) | 42 (0.30) | −0.038 | −0.005 |
| Bowel preparation quality | ||||||
| Excellent | 8022 (28.1) | 8053 (33.1) | 1062 (23.4) | 3484 (24.8) | −0.407 | 0.010 |
| Good | 18617 (65.2) | 14912 (61.4) | 2984 (65.6) | 10013 (71.4) | −0.864 | −0.014 |
| Fair | 1924 (6.7) | 1343 (5.5) | 501 (11.0) | 534 (3.8) | −0.012 | 0.009 |
Abbreviations: BMI, body mass index, NSAID, non-steroidal anti-inflammatory
Sample restricted to exams performed by endoscopists with 5%-95% propofol use at facilities with 5%-95% propofol use
Standardized differences in patient characteristics before and after inverse probability of treatment weighting, sample restricted to exams performed by endoscopists with 5%-95% propofol use at facilities with 5%-95% propofol use
Variables where imputation was used
Regular: once or more per week
Percent missing, Full sample: moderate sedation/propofol: BMI: 17.0% /19.3%; family history of colorectal cancer in 1st degree relative: 15.3%/13.0%; aspirin use: 15.0%/12.5%; NSAID use: 13.1%/10.5%; blood thinner use: 8.1%/7.0%; smoking status: 11.1%/9.2%; alcohol use: 11.4% /9.5%; general health: 11.2%/9.6%; insurance: 4.0%/3.1%; bowel preparation quality: 2.6%/1.8%.
Percent missing: Restricted sample: BMI: 16.8%/19.2%; moderate sedation/propofol: family history of colorectal cancer in 1st degree relative: 16.2% /13.7%; aspirin use: 15.0%/12.4%; NSAID use: 12.9%/10.3%; blood thinner use: 7.7%/6.4%; smoking status: 11.0%/8.9%; alcohol use: 10.9%/9.2%; general health: 10.6%/9.2%; insurance: 4.0%/3.4%; bowel preparation quality: 2.4%/2.1%.
In both cohorts, the majority of exams comprised screening exams (propofol: 64.1% screening and 35.9% surveillance; moderate sedation: 66.8% screening and 33.3% surveillance) and few examinations had a quality of bowel preparation classified as fair (6.7% vs 5.5%). Both the propofol and moderate sedation groups had a similar incidence of family history of colorectal cancer (23.5% vs 24.7%) as well as previous examinations where neoplastic lesions were identified (37.7% vs 36.5%). Our restricted sample included 18,998 examinations performed by endoscopists at facilities with varying propofol use, again with similar proportions of screening vs surveillance exams in both cohorts (surveillance indication: 42.9% propofol and 37.9% moderate sedation). In the restricted sample, the standardized mean differences before and after re-weighting the sample by the inverse probability of treatment (i.e. the propensity score) reflected a balanced sample with no differences in the demographic, clinical and procedural characteristics between the cohorts based on a standardized mean difference threshold of 0.1 (Table 1).
Polyp prevalence for each cohort is presented in Figures 2a (full sample) and 2b (restricted sample). See table, Supplemental Digital Content 6, for the corresponding data for each figure reporting polyp prevalence. In our full sample analysis, 58.1% (CI 57.6–58.7) of exams utilizing propofol detected at least one neoplastic lesion whereas 51.8% (CI 51.2–52.5) of moderate sedation exams detected a neoplastic lesion. There was a small difference in adenoma detection between the groups [39.5% (38.9–40.0) propofol vs. 37.3% (36.7–38.0) moderate sedation]. However, there was a marked difference in serrated polyp detection with propofol use [34.0% (33.4–34.5) propofol vs. 24.5% (24.0–25.1) moderate sedation] (Figure 2a). This pattern of increased serrated polyp detection with propofol use persisted in our restricted sample [30.3 (28.9–31.6) propofol vs. 25.7% (25.0–26.5) moderate sedation] (Figure 2b).
Figure 2a:
Polyp detection by sedation type - full sample1
1 for confidence interval bounds, see Supplemental Digital Content 6a
Figure 2b:
Polyp detection by sedation type - restricted sample1
1 for confidence interval bounds, see Supplemental Digital Content 6b
Table 2 shows the results from our full-sample regression analyses. The first row lists the unadjusted odds ratio (OR) (95% confidence interval) for propofol use vs moderate sedation for the detection of any neoplasia, OR 1.29 (1.25–1.33), adenoma, OR 1.09 (1.05–1.13), and serrated lesion, 1.57 (1.51–1.63). The second row contains results adjusted for several potentially confounding variables in the multivariate logistic regression model: age, body mass index, patient sex, alcohol use, smoking status, self-reported health, family history of colorectal cancer in a first degree relative, patient history of previous colonoscopy, patient history of previous neoplastic polyp findings, and exam preparation quality. Adjusted odds ratios for polyp detection with propofol were OR 1.25 (1.21–1.29) for any neoplasm, OR 1.07 (1.03–1.11) for adenomas and OR 1.51 (1.46–1.57) for serrated polyps. When we incorporated random effects for facility using a generalized mixed model analysis, ORs for polyp detection with propofol were OR 0.95 (0.89 – 1.02) for any neoplasm, OR 0.98 (0.91 – 1.05) for adenomas and OR 1.02 (0.95 – 1.10) for serrated polyps.
Table 2:
Full sample logistic regression results1 – propofol use odds ratios and 95% confidence intervals
| Any neoplasia2 | Adenoma2 | Serrated polyp2 | |
|---|---|---|---|
| Univariate logit | 1.29 (1.25–1.33) | 1.09 (1.05–1.13) | 1.57 (1.51–1.63) |
| Multivariate logit 3 | 1.25 (1.21–1.29) | 1.07 (1.03–1.11) | 1.51 (1.46–1.57) |
| Multivariate Generalized linear mixed effects models with random effects for facility 3 | 0.95 (0.89 – 1.02) | 0.98 (0.91 – 1.05) | 1.02 (0.95 – 1.10) |
Sample includes all exams meeting inclusion criteria, regardless of endoscopist (facility average propofol use;
Odds ratio and 95% confidence interval
Variables included in multivariate regressions are age, BMI, patient sex, alcohol use, smoking habit, self-reported health, family history of colorectal cancer in first degree relative, patient history of previous colonoscopies, patient history of previous neoplastic polyp findings, and bowel preparation quality.
In our restricted sample analysis limited to endoscopists and facilities with variable propofol use, the unadjusted odds ratio for detection with propofol was OR 1.06 (0.99–1.13) for any neoplasm, OR 0.99 (0.93–1.06) for adenomas and OR 1.25 (1.16–1.35) for serrated polyps (Table 3). Following the inverse probability treatment weighting adjustment described in Table 1, increased OR for serrated polyps in the propofol cohort persisted; OR 1.16 (1.11–1.22) and ORs on propofol vs. moderate sedation were not statistically significant for adenomas, OR 0.89 (0.86–0.93) and for any neoplasia, OR 0.95 (0.91–0.99). With adjustment for endoscopist random effects using a generalized linear mixed effects model, univariate (no propensity score adjustment) ORs for polyp detection with propofol were 1.15 (1.06 – 1.24) for any neoplasm, 1.11 (1.03–1.20) for adenomas, and 1.18 (1.09 – 1.29) for serrated polyps, while inverse probability of treatment weighted ORs for polyp detection with propofol were 1.03 (0.98–1.08) for any neoplasm, 1.00 (0.95–1.05) adenomas and 1.13 (1.07–1.19) for serrated polyps (table 3). Adjusting for time in years in the inverse probability of treatment weighted model produced similar results. See table, Supplemental Digital Content 7, for the results of the restricted sample analysis where time is treated as a categorical variable in our sensitivity analysis. Propensity score category results were similar except that with random effects there was a slightly smaller effect size for serrated polyps (OR 1.10 (1.01–1.20)). Refer to figures in Supplemental Digital Content 8a–8h for caterpillar plots illustrating the correlation between propofol use and the detection of any neoplasm, adenomas and serrated polyps by endoscopists and facilities.
Table 3:
Restricted sample logistic regression results1 – propofol use odds ratios and 95% confidence intervals
| Regression type | Propensity score adjustment | Any neoplastic lesion2 | Adenoma2 | Serrated polyp2 |
|---|---|---|---|---|
| No random effects | Univariate – no adjustments | 1.06 (0.99–1.13) | 0.99 (0.93–1.06) | 1.25 (1.16–1.35) |
| Propensity score categories | 0.95 (0.89–1.02) | 0.89 (0.83–0.96) | 1.15 (1.07–1.24) | |
| Inverse probability of treatment weighted | 0.95 (0.91–0.99) | 0.89 (0.86–0.93) | 1.16 (1.11–1.22) | |
| Generalized linear mixed effect models with endoscopist random effects | Univariate – no adjustments | 1.15 (1.06–1.24) | 1.11 (1.03–1.20) | 1.18 (1.09–1.29) |
| Propensity score categories | 1.03 (0.95–1.12) | 1.00 (0.92–1.08) | 1.10 (1.01–1.20) | |
| Inverse probability of treatment weighted | 1.03 (0.98–1.08) | 1.00 (0.95–1.05) | 1.13 (1.07–1.19) |
Sample restricted to exams performed by endoscopists with 5%-95% propofol use at facilities with 5%-95% propofol use;
Odds ratio and 95% confidence interval
Propensity scores were obtained by modeling the probability of being administered propofol using a logistic regression including terms for: age, BMI, sex, alcohol use, smoking status, self-reported health, first degree relative with history of colorectal cancer, previous colonoscopy, history of previous neoplastic lesions, and bowel preparation quality
DISCUSSION
In our statewide sample of screening and surveillance colonoscopies, we observed that propofol sedation when compared to moderate sedation without propofol was associated with a higher likelihood of serrated polyp detection. We did not observe significant differences in total polyp detection, or in the detection of adenomas.
The few studies investigating the relationship between propofol administration and polyp detection have been conflicting.9–14 In four studies, propofol was not associated with improvement in precancerous lesion identification.11–13,18 Wang et. al. found that propofol was associated with increased detection of neoplasms ≥10 mm, but did not influence detection of smaller lesions. However, only 3% of subjects received propofol and histology was not used to categorize polyp type.14 Other studies have identified improved polyp detection with propofol. Using a cohort of over 44,000 patients, Abu Baker, et al. discovered, after adjusting for confounders, that propofol was associated with increased cecal intubation and polyp detection.9 However, unlike the current data, their investigation included patients where colonoscopy was indicated for inflammatory bowel disease, rectal bleeding, and potential malignancy rather than focusing on screening colonoscopies. In an investigation of the impact of anesthesia on missed neoplasms, Dong et al. determined that polyp miss rates were 9.2% lower with propofol, though this study was limited by small sample size.10 Xu et al identified that colonoscopies performed with sedation were associated with higher adenoma detection rates, a quality indicator measuring endoscopist performance. However, they did not specify the sedation type evaluated in their analysis.19
An important distinction of our study is that the analysis leverages the population-based New Hampshire Colonoscopy Registry, which includes data on essential clinical quality and procedural measures, such as adequacy of bowel preparation, detailed polyp histology, polyp location, size, and exam completion status since 2004. This level of detail allows for multivariable regression analyses to adjust for important covariates while maintaining robust restricted samples.17 Another distinctive feature is that this registry obtains histological reports directly from pathology labs, accurately linking polyp-level pathology information to each procedure within the database. This allows us to distinguish between polyps based on specific histological classifications. In addition, detailed tracking allows for endoscopist and facility level analyses and adjustments where relevant. The New Hampshire Registry has played a pivotal role in improving the quality of colorectal cancer screening- including providing evidence that raised the benchmark standards for colonoscopy withdrawal times to improve adenoma and serrated polyp detection.20
Adenomas comprise about 70% of precancerous polyps detected during colonoscopy.4,5 Serrated polyps tend to be flatter, blending into colonic folds, and often covered with a mucus cap, and therefore can be more challenging to detect and remove than adenomas.6–8 Pooling adenomas with serrated lesions for analysis may lead to the erroneous conclusion of no measurable difference in overall polyp detection with propofol use. In this study, we discovered that propofol-based deep sedation was associated with higher detection of the serrated lesions, supporting the hypothesis that propofol may have differential effect in improving the detection of more difficult to identify lesions.
The serrated neoplastic pathway has been increasingly recognized for its association with a higher incidence of interval cancers, which are colorectal cancers detected within a short timeframe after colonoscopy, but before the next recommended examination. The similarity in histological features between serrated polyps and interval cancers highlights the need for interventions aimed at optimizing their detection.
Our findings persisted in sensitivity and alternative analyses which incorporated adjustment for known confounders, clustering of patients within facilities or endoscopists, and focused on facilities with variable use of propofol sedation. The effect sizes for detection of serrated polyps using propofol sedation compared to moderate sedation ranged from to 1.02 (0.95 – 1.10) to 1.51 (1.46 – 1.57). In a restricted sample of endoscopists and facilities that used propofol for between 5% and 95% of procedures, our generalized linear mixed effects model with endoscopist random effects takes into consideration differences in endoscopist skill level, years of experience, variations in technique, and residual unmeasured factors which could independently influence polyp detection. Across all restricted sample analyses, the effect size of serrated polyp detection with propofol use remained higher in comparison to the effect size for the detection of any neoplasm or of adenomas. In this restricted sample analysis, the effect size for detection of serrated polyps using propofol sedation ranged from 1.13 (1.07 – 1.19) to 1.25 (1.16 – 1.35). Restricting our analysis to endoscopists with variable propofol use decreased the sample size by over 50% (down to 18,998 exams) and was weighted towards fewer examinations in the propofol cohort (25% in the restricted sample vs 54% in the full sample). Eliminating sites that exclusively used propofol resulted in a smaller observed effect size, but the relative relationship between serrated polyp detection compared to other polyp type detection was consistent.
Since propofol use is inextricably tied to those practice sites that exclusively used propofol, we were unable to draw meaningful conclusions from our full sample generalized linear mixed model including random effects for facility (Table 2), as for many facilities, this was tantamount to controlling for our exposure of interest. The majority of exams in our full-sample analysis were performed at practice sites/facilities with very little variation in propofol use - in other words, sites that either almost never use propofol or that use propofol for nearly every exam. See figures, in Supplemental Digital Content 8a and 8b of caterpillar plots documenting the relationship between propofol use by practice site and endoscopist, respectively. It is therefore unsurprising that controlling for the influence of site by including site-level random effects in our generalized linear mixed model did not identify a statistically significant association between propofol use and overall polyp detection in the full sample.
Our findings that propofol-based deep sedation is associated with a higher likelihood of serrated polyp detection have important implications for screening and surveillance colonoscopy. Although polyps are more common in men, women over the age of 50 are more likely to develop pre-malignant lesions in the proximal colon, where serrated lesions are more common.21 As research continues to evolve, identifying patient populations at greatest risk for serrated polyp development and targeted strategies aimed at improving detection of these high-risk lesions will be essential. Our study provides preliminary evidence that propofol-based sedation may play an important role in ensuring optimal colonoscopy polyp detection. Additional studies to clarify the effectiveness of propofol will guide understanding of best practices for colorectal cancer prevention.
Our findings have several limitations. A significant limitation is that propofol mediated sedation is typically administered based on site and endoscopist preference, as well as availability of anesthesia staff, rather than being randomly assigned. Thus, without a randomized controlled trial we cannot exclude the possibility of continued residual confounding. Ultimately, a prospective trial where patients are randomly assigned to receive propofol should be considered for future study. Another significant limitation is that due to the nature of our sample, we were unable to adjust our full sample analysis at both the center and the endoscopist level. The reason for this is that the full sample regression analyses did not reliably converge when using random effects for both facility and endoscopist.
Next, although our study comprised 30 geographically, ethnically, and socioeconomically distinct practices, our population was limited to those receiving care in New Hampshire, thus raising the possibility that our study may not be generalizable to areas outside of northern New England.22 Although our generalized linear mixed effects model with endoscopist random effects incorporated measured and unmeasured differences in endoscopist characteristics such as skill level, we did not formally determine the relationship between endoscopist adenoma detection rate and serrated detection rate, two quality indicators known to be associated with increased polyp detection.23 In addition, the most robust models in the restricted sample, which also incorporated patient clustering at the endoscopist level, demonstrated much smaller effect sizes than the overall sample analyses; an effect size of 1.10 or 1.13 may reflect residual confounding.
In conclusion, given the significant challenges inherent to serrated polyp detection, and the fact that missing these lesions undermines the effectiveness of colonoscopy in colorectal cancer prevention, strategies are needed to improve their identification. As propofol use for colonoscopy continues to grow nationally, there is meaningful debate regarding the effectiveness of this practice. Our study provides evidence that propofol use may be associated with increased serrated polyp detection; since these lesions may be the precursors for up to 30% of colorectal cancers, improving their detection may be of substantial benefit.
Supplementary Material
Supplemental Digital Content 8: Correlation between propofol use and detection of any neoplasms, adenomas and serrated polyps by endoscopists and facilities
Supplemental Digital Content 2: New Hampshire colonoscopy registry data dictionary
Supplemental Digital Content 1: New Hampshire colonoscopy registry data collection protocol
Supplemental Digital Content 3: Model specifications and covariates for the full sample
Supplemental Digital Content 4: Model specifications and covariates for the restricted sample
Supplemental Digital Content 5: Model adjustments and rationale
Supplemental Digital Content 6: Polyp detection by sedation type in full and restricted sample
Supplemental Digital Content 7: Restricted sample logistic regression with sensitivity analysis incorporating year as a categorical variable
Acknowledgments
Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health under Award Number R01CA243449. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
Clinical trial number: Not applicable
Propofol-Based Sedation during Colonoscopy Associated with Increased Serrated Polyp Detection; presented at the Association of University Anesthesiologists Annual Meeting on April 13, 2023 in Denver Colorado and at the Costas T. Lambrew Research Retreat on May 3, 2023 in Portland, ME.
References
- 1.American Cancer Society. Cancer Facts & Figures 2021. Accessed 12/20/2023. https://www.cancer.org/content/dam/cancer-org/research/cancer-facts-and-statistics/annual-cancer-facts-and-figures/2021/cancer-facts-and-figures-2021.pdf. [Google Scholar]
- 2.World Health Organization. Cancer. Accessed 12/20/2023. https://www.who.int/news-room/fact-sheets/detail/cancer. [Google Scholar]
- 3.Xi Y, Xu P: Global colorectal cancer burden in 2020 and projections to 2040. Transl Oncol 2021; 14: 101174. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Bond JH: Polyp guideline: diagnosis, treatment, and surveillance for patients with colorectal polyps. Practice Parameters Committee of the American College of Gastroenterology. Am J Gastroenterol 2000; 95: 3053–63 [DOI] [PubMed] [Google Scholar]
- 5.Risio M: The natural history of adenomas. Best Pract Res Clin Gastroenterol 2010; 24: 271–80 [DOI] [PubMed] [Google Scholar]
- 6.Anderson JC, Srivastava A: Colorectal Cancer Screening for the Serrated Pathway. Gastrointest Endosc Clin N Am 2020; 30: 457–478 [DOI] [PubMed] [Google Scholar]
- 7.Crockett SD, Nagtegaal ID: Terminology, Molecular Features, Epidemiology, and Management of Serrated Colorectal Neoplasia. Gastroenterology 2019; 157: 949–966.e4 [DOI] [PubMed] [Google Scholar]
- 8.Rex DK, Ahnen DJ, Baron JA, Batts KP, Burke CA, Burt RW, Goldblum JR, Guillem JG, Kahi CJ, Kalady MF, O’Brien MJ, Odze RD, Ogino S, Parry S, Snover DC, Torlakovic EE, Wise PE, Young J, Church J: Serrated lesions of the colorectum: review and recommendations from an expert panel. Am J Gastroenterol 2012; 107: 1315–29; quiz 1314, 1330 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Abu Baker F, Mari A, Aamarney K, Hakeem AR, Ovadia B, Kopelman Y: Propofol sedation in colonoscopy: from satisfied patients to improved quality indicators. Clin Exp Gastroenterol 2019; 12: 105–110 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Dong H, Ren Y, Jiang B: Risk factors associated with missed colorectal lesions in colonoscopy and impact of colonoscopy with anesthesia on miss rate. Scand J Gastroenterol 2021; 56: 484–491 [DOI] [PubMed] [Google Scholar]
- 11.Metwally M, Agresti N, Hale WB, Ciofoaia V, O’Connor R, Wallace MB, Fine J, Wang Y, Gross SA: Conscious or unconscious: the impact of sedation choice on colon adenoma detection. World J Gastroenterol 2011; 17: 3912–5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Nakshabendi R, Berry AC, Munoz JC, John BK: Choice of sedation and its impact on adenoma detection rate in screening colonoscopies. Ann Gastroenterol 2016; 29: 50–5 [PMC free article] [PubMed] [Google Scholar]
- 13.Paspatis GA, Tribonias G, Manolaraki MM, Konstantinidis K, Chainaki I, Theodoropoulou A, Vardas E, Chlouverakis G: Deep sedation compared with moderate sedation in polyp detection during colonoscopy: a randomized controlled trial. Colorectal Dis 2011; 13: e137–44 [DOI] [PubMed] [Google Scholar]
- 14.Wang A, Hoda KM, Holub JL, Eisen GM: Does level of sedation impact detection of advanced neoplasia? Dig Dis Sci 2010; 55: 2337–43 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP: The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 2008; 61: 344–9 [DOI] [PubMed] [Google Scholar]
- 16.Preventive US Services Task Force, Davidson KW, Barry MJ, Mangione CM, Cabana M, Caughey AB, Davis EM, Donahue KE, Doubeni CA, Krist AH, Kubik M, Li L, Ogedegbe G, Owens DK, Pbert L, Silverstein M, Stevermer J, Tseng CW, Wong JB: Screening for Colorectal Cancer: US Preventive Services Task Force Recommendation Statement. JAMA 2021; 325: 1965–1977 [DOI] [PubMed] [Google Scholar]
- 17.Anderson JC, Hisey W, Mackenzie TA, Robinson CM, Srivastava A, Meester RGS, Butterly LF: Clinically significant serrated polyp detection rates and risk for postcolonoscopy colorectal cancer: data from the New Hampshire Colonoscopy Registry. Gastrointest Endosc 2022; 96: 310–317 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Thirumurthi S, Raju GS, Pande M, Ruiz J, Carlson R, Hagan KB, Lee JH, Ross WA: Does deep sedation with propofol affect adenoma detection rates in average risk screening colonoscopy exams? World J Gastrointest Endosc 2017; 9: 177–182 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Xu C, Tang D, Xie Y, Ni M, Chen M, Shen Y, Dou X, Zhou L, Xu G, Wang L, Lv Y, Zhang S, Zou X: Sedation Is Associated with Higher Polyp and Adenoma Detection Rates during Colonoscopy: A Retrospective Cohort Study. Gastroenterol Res Pract 2023; 2023: 1172478. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Butterly LF: Withdrawal Time: Is Nine the New Six? Clin Gastroenterol Hepatol 2022; 20: e22–e24 [DOI] [PubMed] [Google Scholar]
- 21.Kim SE, Paik HY, Yoon H, Lee JE, Kim N, Sung MK: Sex- and gender-specific disparities in colorectal cancer risk. World J Gastroenterol 2015; 21: 5167–75 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Rice K, Gressard L, DeGroff A, Gersten J, Robie J, Leadbetter S, Glover-Kudon R, Butterly L: Increasing colonoscopy screening in disparate populations: Results from an evaluation of patient navigation in the New Hampshire Colorectal Cancer Screening Program. Cancer 2017; 123: 3356–3366 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Anderson JC, Butterly LF, Weiss JE, Robinson CM: Providing data for serrated polyp detection rate benchmarks: an analysis of the New Hampshire Colonoscopy Registry. Gastrointest Endosc 2017; 85: 1188–1194 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental Digital Content 8: Correlation between propofol use and detection of any neoplasms, adenomas and serrated polyps by endoscopists and facilities
Supplemental Digital Content 2: New Hampshire colonoscopy registry data dictionary
Supplemental Digital Content 1: New Hampshire colonoscopy registry data collection protocol
Supplemental Digital Content 3: Model specifications and covariates for the full sample
Supplemental Digital Content 4: Model specifications and covariates for the restricted sample
Supplemental Digital Content 5: Model adjustments and rationale
Supplemental Digital Content 6: Polyp detection by sedation type in full and restricted sample
Supplemental Digital Content 7: Restricted sample logistic regression with sensitivity analysis incorporating year as a categorical variable