Is Cognitive Impairment Associated with Inadequate Bowel Preparation for Colonoscopy?

Abstract

Purpose:

Inadequate bowel preparation (IBP) for colonoscopies is associated with missed polyps and cancers, prolonged procedure times, lower colonoscopy completion rates, and need for a repeat procedure. The purposes of this study were to assess: 1) whether impaired cognitive status (measured by an abnormal clock drawing test) was associated with IBP, and 2) the association of participant demographic and clinical characteristics with IBP.

Methods:

Cross-sectional cohort study conducted in 3 academic centers. Subjects, 50 to 85 years old, completed 5 stool tests on a single sample and a clock drawing before undergoing a screening or surveillance colonoscopy. Clock drawings were validated by the Mendes-Santos method and Watson method. A generalized linear mixed model was conducted to estimate factors associated with IBP, based on Aronchick bowel preparation score.

Results:

The 2,016 participants had a mean age of 63 years (SD = 7.95); 1,274 (63%) were female; 119 (6%) had IBP; and 421 (21%) had an abnormal clock drawing. After controlling for age in the multivariable model, the odds of having an IBP versus adequate were 1.44 (95% CI, 0.94–2.22) for those with an abnormal clock drawing score compared with those with a normal score. The only variable significantly associated with IBP was household income, with an odds of 2.48 (95% CI, 1.56–3.95) for household income of <$40,000 compared with income >$80,000.

Conclusions:

Lower household income was associated with greater odds of IBP. The association between an abnormal clock drawing score and IBP was not statistically significant.

NCT03264898 (clinicaltrials.gov) – Comparative Effectiveness of FITs with Colonoscopy

Colonoscopy is regarded as the standard and the most effective tool for colorectal cancer screening, and as such, it is important for adequate bowel cleaning.^1,2 Rates of inadequate bowel preparation (IBP) for colonoscopies range from 9% to 48% with 2 of the higher percentages of IBP for those who are obese.^3–16 IBP is associated with missed neoplasia/diagnosis, prolonged procedure times, lower cecum intubation rates, and need for repeat procedure.^17–23 Prevalence of missed adenomas range from 22% to 42%.^3,9,24–26 A number of factors have been associated with IBP, including certain demographic characteristics (high school education or less, lower socioeconomic status, marital status, minority),^{7,9,10,13,17,27–29} history of prior abdominal surgeries,^30–32 medical diagnoses (obesity, cirrhosis, diabetes, hypertension, and multiple sclerosis),^4,8,10,12,33 and medications (antipsychotics/antidepressants, calcium channel blockers, NSAIDs, opioids, and polypharmacy).^4,5,10,11,27 In addition, prior retrospective studies have linked decreased mental capacity (degree of understanding and memory), Parkinson disease, and stroke to risk of IBP.^4,8,10

Standard guidelines for colonoscopy follow-up intervals are based on persons having excellent or good bowel preparation.² However, endoscopists often recommend that persons with IBP return earlier with more intensive or prolonged bowel preps. Clark et al. conducted a systematic review and determined that the adenoma detection rate is significantly higher for those with a high-quality bowel preparation compared with those with a low-quality preparation.² Patients who received additional education regarding bowel preparation have significantly better bowel preparation.^34–36

The literature is limited on associations between cognitive impairment and IBP. We previously found that about a quarter of subjects participating in a study to compare the effectiveness of 5 different fecal immunochemical tests (FITs) with optical colonoscopy had an abnormal clock drawing, which was associated with FIT collection errors.³⁷ The purposes of our current study were to assess: 1) whether impaired cognitive status (measured by abnormal clock drawing test) was associated with IBP, and 2) the association of participant demographic and clinical characteristics with IBP.

Methods

The parent study prospectively evaluated the diagnostic performance of 5 FIT tests in patients scheduled for a screening or surveillance colonoscopy. A detailed description of the study is provided elsewhere.^38–40 Briefly, inclusion criteria were individuals aged 50 to 85 years and able to speak, read, and write in English or Spanish. We excluded persons with iron deficiency anemia, a personal history of colorectal cancer, inflammatory bowel disease, ulcerative colitis or Crohn's disease, rectal bleeding in the previous 2 months, personal or family history of a familial adenomatous polyposis, hereditary nonpolyposis colon cancer, or the colonoscopy being recommended for diagnostic purposes, and dementia.³⁹ Institutional Review Board approval was received for this study at the 3 participating sites: University of Iowa (UI) in Iowa City, Iowa; University of North Carolina (UNC) at Chapel Hill, North Carolina; and Texas Tech University Health Sciences Center (TTUHSC) at El Paso, Texas.

Each subject in this analysis returned a health questionnaire and product questionnaire (developed by the investigators), completed FIT tests, and underwent colonoscopy. Potential participants were invited to participate, and after the informed consent and health questionnaire were returned, they were provided a box of supplies which included instructions for collecting the stool specimens and returning them, 5 FITs, a specimen collection container, a pair of gloves, a yellow card to record the date of stool collection, a small cardboard postage-paid return mailer, and a product questionnaire with postage-paid return envelope. The product questionnaire included questions regarding the ease of collection of the stool for each product and a clock drawing test.

Health Questionnaire

The 19-item health questionnaire was adapted from a previous study.⁴¹ Demographic and clinical questions included: date of birth, sex, race, ethnicity, education, household income, height, weight, and zip code. Smoking status was self-reported as current, former, or never smoked and dichotomized as currently smoking yes or no.

Clock Drawing Test

Subjects were instructed to draw the face of a clock using a preprinted circle (“Please draw the face of a clock with numbers and hands to signify the time Ten minutes after eleven”) and to indicate if anyone helped them draw the clock. Each clock was scored using the Watson et al. and Mendes-Santos et al. methods.^37,42,43

The Watson et al. scoring method involved dividing the circle of a clock into 4 quadrants and drawing a line through the center of the circle and the number 12.⁴² Then, a second line is drawn that is perpendicular to and bisecting the first line resulting in 4 quadrants. The Watson et al. method does not consider the position of the hour nor the minute hand. The best possible score is 0 and the worst possible score is 7. Watson et al., defined scores of 0–3 as normal cognition and scores of 4–7 as abnormal cognitive function.⁴²

The second scoring method, the Mendes-Santos et al. method, was adapted from the Sunderland et al. method.^43,44 The Mendes-Santos et al. method included a list of 17 items identified as present or absent. Scores can range from 1 to 10 with higher scores indicating better performance. For the Mendes-Santos et al. method, scores 6–10 were classified as normal and score 1–5 were abnormal.⁴³

Primary Outcome Variable

Colonoscopy reports were obtained from the subject’s electronic medical record. A colonoscopy/pathology review form^38,41,45 was used to record the colonoscopist’s description of the bowel preparation. The primary outcome variable was bowel preparation quality, recategorized according to the Aronchick scale: excellent, good/adequate, fair, poor, or inadequate. Based on these descriptions, we characterized the bowel preparation quality as inadequate (fair, poor, or inadequate) or adequate (excellent, good, or adequate).^30,46

Statistical Analysis

Standard descriptive statistics were summarized for each variable: continuous variables as mean and standard deviation and categorical variables as frequency and percentage.

A generalized linear mixed model using the SAS GLIMMIX procedure was used to estimate a binomial model predicting the outcome variable of whether a participant’s bowel preparation quality was inadequate vs adequate. The site variable (Iowa, North Carolina, or Texas) was specified as the random intercept in the model to account for the correlation among participants within each site. First, each of the potential associated factors was tested separately with the outcome variable in the model. Variables tested included participants’ age, body mass index (BMI), sex, race, ethnicity, education, household income, smoking status, unevaluable FITs (any one FIT from a participant unevaluable), colonoscopy type (surveillance or screening), and each of the 2 clock drawing scores (abnormal vs normal). Variables with P < .15 in the univariable predictor models were included in the multivariable model analyses. P values <0.05 were considered statistically significant. All analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC).

Results

Of the total 2,074 participants who completed the health questionnaire, FITs, product questionnaire, and colonoscopy procedure, 37 did not draw the clock, 17 received help on clock drawing, and 4 did not have quality of bowel preparation documented by the colonoscopist, leaving 2,016 participants left for analysis. Of the 2016 participants, 119 (5.9%) had an IBP, 421 (20.9%) had abnormal clock drawings by the Mendes-Santoset al method, and 537 (26.6%) had abnormal clock drawings by the Watson et al. method. Seven hundred forty-two (36.8%) of the participants were male, 1,687 (83.7%) were White, 381 (18.9%) were Hispanic, with a mean age of 62.8 years. Seventy-eight (3.9%) had an eighth-grade education or less and 572 (28.4%) had an income of less than $40,000. One hundred and 6 (5%) participants were smokers and 75 (4%) had at least one unevaluable FIT. Participants’ mean BMI was 28.8 (See Table 1).

Table 1.

Participant Characteristics by Bowel Preparation Quality (n = 2,016)

	Inadequate Preparation	Adequate Preparation
	(±SD) or n (%)	(±SD) or n (%)
Characteristics	Total n = 119	Total n = 1,897
Age, years
Mean (±SD)	61.5 (±7.4)	62.9 (±8.0)
Age
<60	52 (7.0)	695 (93.0)
≥60.0	67 (5.3)	1,202 (94.7)
BMI (±SD)	29.8 (±7.8)	28.7 (±6.4)
BMI
<30.0	73 (5.6)	1,233 (94.4)
≥30.0	45 (6.4)	662 (93.6)
Site
UI	76 (9.1)	761 (90.9)
UNC	33 (4.2)	749 (95.8)
TTUHSC	10 (2.5)	387 (97.5)
Sex
Female	68 (5.3)	1,206 (94.7)
Male	51 (6.9)	691 (93.1)
Race
Black	10 (7.2)	129 (92.8)
White	100 (5.9)	1,587 (94.1)
Others	9 (4.7)	181 (95.3)
Ethnicity
Hispanic	11 (2.9)	370 (97.1)
Non-Hispanic/unknown	108 (6.6)	1,527 (93.4)
Education
≤8th grade	3 (3.9)	75 (96.2)
HS/GED	25 (6.5)	359 (93.5)
College or higher	90 (5.9)	1,447 (94.1)
Not reported	1 (5.9)	16 (94.1)
Income
<$40,000	47 (8.2)	525 (91.8)
$40,000–<$80,000	21 (4.6)	436 (95.4)
≥$80,000	47 (5.2)	858 (94.8)
Not reported	4 (4.9)	78 (95.1)
Current Smoker
Yes	13 (12.3)	93 (87.7)
No	106 (5.6)	1,796 (94.4)
Unevaluable FIT
Yes	2 (2.7)	73 (97.3)
No	117 (6.0)	1,824 (94.0)
Colonoscopy type
Screening	74 (5.6)	1,252 (94.4)
Surveillance	45 (6.5)	645 (93.5)
Clock drawing test Watson
Abnormal	37 (6.9)	500 (93.1)
Normal	82 (5.5)	1,397 (94.5)
Clock drawing test Mendes
Abnormal	32 (7.6)	389 (92.4)
Normal	87 (5.5)	1,508 (94.6)

Abbreviations: SD, standard deviation; HS, high school; GED, General Equivalency Development; FIT, fecal immunochemical test; UI, University of Iowa; UNC, University of North Carolina; TTUHSC, Texas Tech University Health Sciences Center.

Table 2 presents the factors associated with an IBP in generalized linear-mixed models. The random site effect in the model indicated the likelihood of IBP was significantly different across 3 sites. In the univariable analyses, being a smoker, and having income less than $40,000 compared with income ≥ $80,000 were significantly associated with IBP. Neither the dichotomized Watson et al. clock score nor the Mendes-Santos et al. clock score were significantly associated with IBP. In the multivariable model, the only factor statistically significantly predicting IBP was income < $40,000 versus ≥ $80,000 (adjusted odds ratio [aOR] 2.48, 95% CI, 1.56-3.95). Age less than 60 years compared with ≥ 60 years (aOR 1.38, 95% CI, 0.94-2.03), current smoker (aOR 1.88, 95% CI, 0.98-3.62), and abnormal Mendes-Santos score (aOR 1.44, 95% CI, 0.94-2.22) did not reach the prespecified statistical significance levels after controlling for other variables in the model (See Table 2). None of the interactions tested for were significant, including age with income, age with the Mendes-Santos clock score, and income with the Mendes-Santos clock score.

Table 2.

Generalized Linear Mixed Model of Factors Associated with Colonoscopy Fair/Poor/Inadequate vs Excellent/Good/Adequate Preparation (n = 2,016)

Factors	Univariable Model		Multivariable Model
	Odds Ratio (95% CI)	P value	Odds Ratio (95% CI)	P value
Age
<60	1.41 (0.97-2.07)	0.075	1.38 (0.94-2.03)	0.11
≥60.0	Reference		Reference
BMI
<30.0	0.86 (0.60-1.30)	0.536
≥30.0	Reference
Sex
Female	0.81 (0.56-1.19)	0.281
Male	Reference
Race
White	0.71 (0.36-1.42)	0.336
Others	0.78 (0.30-2.06)	0.619
Black	Reference
Ethnicity
Hispanic	0.73 (0.27-1.98)	0.539
Non-Hispanic/unknown	Reference
Education
≤8th grade	1.42 (0.39-5.13)	0.592
HS/GED	1.29 (0.81-2.08)	0.287
College or higher	Reference
Income
<$40,000	2.63 (1.68–4.10)	<0.001	2.48 (1.56-3.95)	<0.001
$40,000–<$80,000	0.91 (0.54-1.55)	0.727	0.95 (0.55-1.63)	0.85
≥$80,000	Reference		Reference
Not reported	1.00 (0.35-2.87)	0.997	1.04 (0.36-2.99)	0.94
Current smoker
Yes	2.57 (1.38-4.78)	0.003	1.88 (0.98-3.62)	0.06
No	Reference		Reference
Unevaluable FIT
Yes	0.44 (0.11-1.80)	0.251
No	Reference
Colonoscopy type
Screening	0.90 (0.61-1.32)	0.580
Surveillance	Reference
Watson method score
Abnormal	1.24 (0.82-1.85)	0.307
Normal	Reference
Mendes method score
Abnormal	1.38 (0.90-2.10)	0.140	1.44 (0.94-2.22)	0.09
Normal	Reference		Reference

Abbreviations: CI, confidence intervals; BMI, body mass index; HS, high school; GED, General Equivalency Development; FIT, fecal immunochemical test.

In the sensitivity analysis, we included 17 participants who received help with clock drawing and categorized them as the abnormal clock drawers. One participant out of 17 had an IBP. We repeated the multivariable model, the association of Mendes-Santos score and IBP remained nonsignificant (aOR 1.43, 95% CI, 0.93-2.19).

Discussion

The clock drawing test, the Mendes-Santos clock drawing score, as a proxy for cognitive status was not associated with IBP. However, we had limited power to detect an effect of cognitive status on IBP, because of the low frequency of inadequate bowel preparation (5.9%) in the analytic sample of 2,016 patients. Abnormal clock drawing was associated with an aOR of 1.44 with a 95% CI of 0.94-2.22. The point estimate of 1.44 and the upper 95% CI of 2.22 would represent a moderate risk factor for inadequate bowel prep. While abnormal clock drawing did not meet our prespecified P-value of <0.05 in the multivariable model, 95 times out of 100, the aOR would fall between 0.94 and 2.22. The lack of precision results from the limited sample size for the uncommon outcome of IBP.

We are unaware of any previous prospective studies that assessed cognitive status as a risk factor for IBP, other than several studies that found those with dementia were more likely to have IBP.^4,5,8 In populations with a high rate of IBP, further study could evaluate if the clock drawing test might be a relatively straightforward method of assessing patients who might need more education regarding their bowel preparation.

Three earlier studies found those with dementia were more likely to have IBP.^4,5,8 In a retrospective review at 4 hospitals on Vancouver Island of 2,101 patient records who underwent a screening or surveillance colonoscopy, diagnoses of dementia/stroke were found to have an odds ratio of 4.2 being associated with IBP.⁴ A cross-sectional study of 649 patients undergoing colonoscopy found history of stroke or dementia were predictors of IBP.⁵ Another study of 2,163 consecutive patients who received outpatient colonoscopies found that those with dementia had IBP.⁸ Our study differs from these 3 studies because we excluded patients with the medical diagnosis of dementia and/or any other diagnosis of memory/cognitive problems. Because of this exclusion criteria, the clock drawing test may have been significant in the model if including all patients having screening/surveillance colonoscopies. For persons over 75 year of age, colorectal cancer screening is generally not recommended.^47,48 Providers should assess the person’s overall health, estimated life expectancy, preferences and values for cancer screening, and the impact of any potential benefits from the cancer screening.⁴⁹ A clock drawing test might be more helpful in an unselected population which could have higher prevalence of dementia and potentially IBP.

For this study, the rate of IBP was quite low at 5.9%, much lower than other studies where IBP rates ranged from 9% to 48%.^{3–17,20,21,27,28,30–33,50–52} This may be partially explained by the relatively high educational level of participants. The other retrospective chart review study in 4 Vancouver Island hospitals which reported a low IBP of 9% did not include race, ethnicity, or education to compare with this study but rather focused on medications, such as opioids, calcium channel blockers, and antidepressants.⁴ Other participant level differences between our study’s participants and prior studies’ participants, were we excluded patients with prior bowel surgeries, dementia, or serious mental health illnesses. Lower education was not associated with IBP in our population, in contrast to other studies where lower education was associated with IBP.^7,20,28,52 We found that lower income was associated with IBP and income is often positively correlated with educational level. Unlike other studies, we found that older age was not predictive of an IBP. Our study is consistent with findings from other studies that lower socioeconomic status,^7,15 and current smoker,^8,10,27 were important predictors of IBP.

Assessing IBP was not the primary purpose of our main study and a limitation of the IBP analysis was that we did not collect data on possible contributing factors such as medical diagnoses, medications, difficulty ambulating, bowel preparation instructions, and medical insurance. Given the low prevalence of IBP in our study population we were likely underpowered to detect a significant association with abnormal clock drawings. The strengths of our study included the large racially and ethnically diverse population attending 3 academic health centers.

Conclusion

Abnormal clock drawing as scored by Mendes-Santos may be an important factor to consider when triaging individuals for a screening/surveillance colonoscopy, given that in our sample, the point estimate of 1.44 and the upper limit of the confidence interval were consistent with it being important. This study had limited power to detect whether abnormal clock drawing was predictive of IBP, due to the low percentage of IBPs in the study. The sample included a diverse population of participants undergoing screening or surveillance colonoscopies at 3 academic medical centers. Future studies could measure cognitive status using the Mendes-Santos et al clock drawing method in settings that have reported a higher prevalence of IBP rates. Given the limited life expectancy of individuals with severe cognitive impairment, these individuals may not be appropriate candidates for screening/surveillance colonoscopies. Identifying at-risk individuals could lead to targeted interventions to reduce IBP, preventing associated risks to patients and avoiding unnecessary use of health care resources.