Accuracy of Self-reported Colonic Polyps: Results from the Prostate, Lung, Colorectal, and Ovarian Screening Trial Study of Colonoscopy Utilization

Colorectal cancer screening is recommended for all adults in the United States aged 50 to 75 years (1, 2). One of the recommended tests, colonoscopy, enables the identification and removal of polyps. Colonoscopy is also used for the diagnostic follow-up of individuals who are screened by other modalities such as fecal immunochemical testing and sigmoidoscopy, and for those with symptoms of colorectal cancer. Finding and removing polyps, more specifically adenomatous polyps (i.e., adenomas), during colonoscopy has implications for subsequent surveillance recommendations, which are guided by the number, size, and histologic type(s) of polyp(s) found (2). Therefore, physicians recommending follow-up surveillance after colonoscopy need to know, with accuracy, the colonoscopy results. This need requires communication between specialty and primary care, and in some cases, the patient may serve as the interface between providers in reporting colonoscopy results (3). Therefore, it is important to understand how accurately patients are able to self-report polyps that are found at colonoscopy.

Self-report of a personal history of polyps is used in other ways as well. For example, some colorectal cancer screening guidelines recommend more intensive screening for individuals who have a family history of adenomatous polyps in a first-degree relative (4). These individuals often rely on the accuracy of family members' reports to inform their own care, making assessments of the validity of self-report of a history of polyps especially important.

To date, few studies have assessed the validity of self-reported colonoscopy findings (5–8), and the studies are further limited by small, nonrepresentative patient populations and/or use of diverse measures of accuracy. To address these gaps, we used data from the Study of Colonoscopy Utilization (SCU), an ancillary study nested within the Prostate, Lung, Colorectal, and Ovarian Screening Trial (PLCO), to determine: (i) how accurately people self-report colonoscopy findings (i.e., whether any polyps or adenomas were discovered), and (ii) whether particular characteristics of the patient or their colonoscopy experience predict accuracy of self-reported colonoscopy findings.

PLCO is a randomized controlled trial that estimated the efficacy of screening for each of the 4 included cancers by comparing cancer-specific mortality between intervention arm participants who were offered screening over a 5-year period and control arm participants who received usual care (9). The trial included men and women 55–74 years of age at enrollment; 77,445 and 77,455 individuals were randomized to the intervention and control arms, respectively. Colorectal cancer screening was initially offered in 1993 as flexible sigmoidoscopy at baseline and again 3 years later, but in 1995 the protocol was changed to offer repeat sigmoidoscopy at 5 instead of 3 years. All participants were asked to respond to an extensive mailed baseline questionnaire that requested information about history of screening for prostate, lung, ovarian, or colorectal cancer, family cancer history, brief medical history, lifestyle behaviors, and demographics (10).

Additional data for this study were obtained from the Study of Colonoscopy Utilization (SCU), an ancillary study nested within the PLCO. A detailed description of SCU can be found elsewhere (11, 12). Briefly, participants were sampled for SCU in varying proportions based on baseline PLCO findings and attendance at year 3/5 repeat sigmoidoscopy. Participants who had more advanced findings at baseline and who did not attend their 3- or 5-year repeat sigmoidoscopy were oversampled. Participants who were selected for and agreed to participate in SCU completed a telephone interview in which they were prompted with the dates of colonoscopies already known to PLCO study staff (i.e., participants were reminded using the phrase, “Our record shows that you had a colonoscopy on [date]”) and asked to report the colonoscopy results, including whether polyps or adenomas were found during each of these known colonoscopies (i.e., participants were asked, “Do you recall if a polyp was found?” and “Do you know what type of polyp was found?”). Response options for the type of polyp found were, “benign, precancerous or adenoma, non precancerous or hyperplastic, and don't know.” In addition, participants were asked to report the occurrence and results of any subsequent colonoscopy occurring after baseline exams that were not already known to SCU staff. Participants reporting a colonoscopy that was not already known to SCU study staff were asked to specify the month and year of the procedure or to provide a general range of when the colonoscopy occurred (<1 year, 1–2 years, 2–3 years, or >3 years). If the PLCO participant was not available to respond, a proxy respondent was invited to participate. Medical and pathological records for all reported colonoscopies were requested from the colonoscopy provider. For colonoscopies for which medical records were located, study staff reviewed the procedure reports and abstracted the presence and type of any polyp(s) found during each colonoscopy.

Of the 4,960 PLCO participants who were selected for and eligible to participate in SCU, 974 were excluded from this study for the following reasons: self-reported or medically confirmed diagnosis of colorectal cancer (n = 22); no reported colonoscopies during SCU (prompted or unprompted, n = 580); no medical record–verified colonoscopies during the SCU study period (n = 76); no medical record for at least one colonoscopy (for those who self-reported more than one colonoscopy during the study period, n = 68), and potential use of a proxy respondent (n = 226 known to use a proxy and n = 2 unknown proxy status). After these exclusions, there were 3,986 participants with at least one self-reported colonoscopy finding that could be compared with medical records (Fig. 1).

As date information provided by a participant reporting a colonoscopy that was not already known to SCU study staff was not always specific, and because a participant could have reported multiple colonoscopies happening in a general time frame (e.g., 2 colonoscopies both 2–3 years ago), it was not possible to link specific self-reported findings to specific colonoscopies in the medical record. Thus, coding of self-report and medical record colonoscopy findings were assessed over the entire study period. From the self-reported data, participants with any self-reported polyp from any colonoscopy during the study period were coded as self-report “yes” for polyp; all other participants were coded as “no.” From the medical records data, participants with any record of a polyp for colonoscopies occurring during the study period were coded as medical record “yes”; all other participants were coded as medical record “no.” Participants were coded similarly for the presence/absence of adenomas.

Descriptive statistics for the overall study population were calculated. Self-reported versus medical record presence of a polyp or adenoma were compared using 2 × 2 tables and by calculating the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for polyp and adenoma separately, with the medical record as the gold standard.

Bivariate and multivariate modeling were used to assess potential predictors of sensitivity and specificity (e.g., accurate self-report separately for people with and without a medical record history of polyp). All logistic regression models were fit with the accuracy of self-report of polyps as the dependent variable (i.e., correctly reported the presence/absence of a polyp). One model was created for those with medical record-verified polyps and another model for those with medical record-verified no polyps. Potential predictors of accurate self-report were chosen a priori and included: the presence of comorbidities (using an index based on 7 comorbid conditions, coded as none, one, or two or more comorbidities), history of diverticulitis/diverticulosis (yes, no), age at SCU participation (60–64, 65–69, 70–74, 75–79, 80+ years of age), race/ethnicity (white non-Hispanic, other), sex (male, female), self-reported family history of colorectal cancer (yes, no), education (≤high school, some college, ≥college degree), time since most recent colonoscopy (0–1 year, 1–<3 years, 3–<5 years, ≥5 years), marital status (married/living as married, not married), smoking status (never, former, or current smoker), total number of colonoscopies that occurred between baseline and SCU participation (1, ≥2), most severe polyp finding (advanced adenoma, nonadvanced adenoma, other polyp, for those with a medical record-verified history of polyp only), and PLCO site (University of Minnesota, Georgetown University, Pacific Health Research and Education Institute (Honolulu), Henry Ford, University of Colorado, Washington University in St Louis, University of Pittsburgh, University of Utah, Marshfield Clinic Research Foundation). We did not estimate a model assessing factors associated with accurate self-report of adenomas because the accuracy of reporting for these was much lower than for polyps.

Bivariate and multivariate logistic regression models produced odds ratios and 95% confidence intervals were calculated using bootstrapping. Specifically, the SAS SURVEYSELECT procedure was used to create 5,000 iterations of bootstrap samples, all capped to the study population size (N = 3,986). Logistic regression models were fit using each bootstrap sampled dataset that produced the distribution of point estimates. All analyses were performed using SAS 9.4 and weighting accounted for the fact that the SCU sample overselected for more “severe” findings from the baseline PLCO colonoscopy and those without a year 3/5 follow-up sigmoidoscopy (11).

A majority of included participants were 70 to 79 years of age (67.6%), female (55.9%), and had no family history of colorectal cancer (87.5%, Table 1). Participants varied in the length of time between their last colonoscopy and participation in the SCU study, with 21.2% having had their colonoscopy in the year prior to the SCU data collection and 16.2% having had their colonoscopy 5 or more years before participating in SCU. A majority of participants only had one colonoscopy during the PLCO study period (60.0%).

Tables 2 and 3 show accuracy of recall for polyp and adenoma. Among the 3,400 people who had a polyp in the medical record, 3,048 self-reported a polyp. Among the 586 people who did not have a polyp recorded in the medical record, 436 self-reported not having a polyp. The weighted sensitivity and specificity for self-reported history of polyp were 0.88 [95% confidence interval (CI): 0.84–0.91] and 0.85 (95% CI: 0.79–0.90), respectively. Among the 2576 people who had an adenoma in the medical record, 380 self-reported an adenoma. Among the 1,410 people who did not have an adenoma in the medical record, 1,393 self-reported not having an adenoma. The weighted sensitivity and specificity for self-reported history of adenoma were 0.11 (95% CI: 0.08–0.15) and 0.99 (95% CI: 0.97–1.00), respectively.

Among participants with a history of polyp (n = 3,400), older age was associated with reduced odds of accurate recall (Table 4). Specifically, compared with participants who were 60–64 at the time of SCU data collection, those 70–74, 75–79, and 80 years or older had 63, 83, and 89 percent lower odds of recalling that they had a previous polyp (OR: 0.37, 95% CI: 0.13–0.91; OR: 0.17, 95% CI: 0.05–0.42; OR: 0.11, 95% CI: 0.03–0.30, respectively). Compared with participants who were white, those who were a race other than white had 8.51 times the odds of accurately self-reporting a polyp (95% CI: 1.05–36.25). People who had a longer time since their last colonoscopy were less likely to accurately self-report a polyp. Specifically, compared with people whose last colonoscopy was in the last year, people whose last colonoscopy was 1–<3 years, 3–<5, or ≥5 years before SCU data collection had 66%, 86%, and 88% lower odds of recalling a previous polyp (OR: 0.34, 95% CI: 0.12–0.92; OR: 0.14, 95% CI: 0.06–0.44; OR: 0.12, 95% CI: 0.04–0.30, respectively). Participants with a finding of advanced or nonadvanced adenoma had 3.72 and 2.36 times the odds of accurately self-reporting a polyp compared with those with other types of polyps (95% CIs: 1.70–7.44 and 1.25–4.72, respectively).

Among those without a history of polyp (n = 586), people who had a longer time since their last colonoscopy were less likely to accurately self-report not having had a polyp (Table 5). Specifically, compared with people whose last colonoscopy was in the past year, people whose last colonoscopy was ≥5 years before SCU data collection had 62% lower odds of accurately recalling not having a previous polyp (OR: 0.38, 95% CI: 0.07–0.83). Having had two or more colonoscopies was associated with 81% lower odds of accurately recalling not having a previous polyp (OR: 0.19, 95% CI: 0.05–0.47). Participants with a family history of colorectal cancer had 4.49 times the odds of accurately recalling not having a polyp compared with those without a family history of colorectal cancer (95% CI: 1.38–17.38).

This study assessed the accuracy of self-reported colonoscopy findings and characteristics associated with accuracy among participants in the PLCO SCU Trial. In this population, participants were able to recall with reasonable accuracy whether they had had a previous polyp during the study period. However, the ability to recall a polyp being an adenoma was very low (sensitivity 11%), even though the ability to recall the absence of an adenoma was nearly perfect (specificity 99%). Possible explanations for these findings include that people might recall having a polypectomy during their colonoscopy but have less ability to recall the pathology results. It is also possible that patients might never have been told the more technical term, adenoma. In this study, adenoma was assessed with the response option, “precancerous or adenoma”; thus, even if a patient had not heard the term adenoma, the alternative definition of precancerous was provided and could have helped to increase self-report of this finding. Future research could use qualitative methods to determine if other terms are being used in doctor-patient discussions of colonoscopy results. It is imperative that medical providers are aware of any polyp findings during a colonoscopy; thus, identifying patient characteristics associated with polyp recall is of importance. Furthermore, by examining correlates of accurate recall separately for people with and without a history of polyps, we were able to identify unique characteristics associated with self-report in both groups.

In this study, increasing time since last colonoscopy was associated with reduced odds of accurate recall for people with and without a history of polyp. These results are similar to Kumaravel and colleagues who found that people who had had a colonoscopy more recently were more likely to have accurate recall of a polyp (7). While perhaps not surprising; these results reinforce the importance of timeliness for patient self-report of colonoscopy findings if they are being used in practice. Among people with a history of polyp, increasing age was associated with reduced likelihood of accurate self-report. Previous studies have not identified age as a negative correlate of recall (5, 8), although one study examining patients' ability to recall surveillance recommendations following a colonoscopy showed increased odds of accurate recall among younger patients (6). The finding of more consequential colonoscopy results predicting increased accuracy among those with a polyp is consistent with Fritz and colleagues, who found that among African American patients, having had an adenoma was associated with increased accuracy of reporting colonoscopy findings (8). Overall accuracy of polyp self-report among their study population was similar to what was found in this study (83.3% accurately recalled their polyp results). Although it has a large total sample, the PLCO SCU was not nationally representative, and just over 8% of the participants were of a race other than white. Even so, multivariate models indicated that non-white participants had increased odds of polyp recall. To our knowledge, only one study has specifically considered accuracy among an African American patient population (8). Future studies should continue to recruit diverse samples to allow for comparisons of self-reported accuracy by race.

Among people without a polyp, family history was associated with increased odds of accurate recall. The association between family history and accurate recall of not having a polyp is unique to this study as previous research to consider this characteristic has not found this association and has also not examined correlates of accuracy separately for people with and without a polyp (6–8). To our knowledge, all previous examinations of accuracy in this context have considered the results of a single colonoscopy. Because of the methods of data collection for the PLCO SCU, we were not able to link all individual colonoscopies to self-report and as a result have considered accuracy for all colonoscopies occurring over the entire study period. This change in consideration of accuracy is perhaps important as, among those without a history of polyps, having had two or more colonoscopies was associated with reduced odds of accurate recall. This association was identified while also accounting for a family history of colorectal cancer and a history of diverticulitis, which could be reasons why someone may have had multiple colonoscopies. Further exploration into the reasons why increased number of colonoscopies reduced recall only for those without polyps is needed to better understand this phenomenon.

Strategies are needed to increase accuracy of patient reporting within primary care. Several studies have examined mechanisms for increasing accurate recall of colonoscopy findings and shown mixed results. One study found that uninsured African American patients who had patient navigation were nine times more likely to accurately report their colonoscopy results than were insured patients under usual care, though those patients were not more likely to accurately recall the appropriate follow-up recommendations (8). Schroy and colleagues found that a computer-based educational program after polypectomy in combination with a personalized letter resulted in improved knowledge of polyp histology compared to the computer program alone or usual care (13). In contrast, another study found that systematic notification of results after a polypectomy, including a verbal report of findings immediately after the procedure, an endoscopy report specifying the findings, and a mailed letter specifying the pathology of the polyps, still resulted in low knowledge of colonoscopy findings among participants (14). Both patient navigation and the delivery of written results sent within close proximity to the colonoscopy have the potential to address reduced recall in specific groups identified in this study, but mixed findings indicate a need to further explore these interventions in additional settings. Furthermore, these interventions would have implications for clinic workload and staffing, which have not been evaluated. However, due to both the low sensitivity of adenoma self-report and decreasing accuracy of self-report over time, encouraging patients to keep colonoscopy and pathology report records and to provide them to new providers who do not have access to this information is likely the most reliable way to ensure that patients receive appropriate surveillance.

This study has several limitations. First, the study entailed a secondary data analysis of the PLCO SCU trial. While this was a large population-based study, as a clinical trial it is possible that the participants in the PLCO SCU were not generalizable to the general population. In addition, not all information of interest was captured in the trial. For example, data were not collected in such a way that self-reported results could be exactly matched to medical records data. Thus, while we were able to assess accuracy over the entire study period, we were not able to assess measures of accuracy by comparing self-reported results to specific medical record results on the basis of individual colonoscopies. However, we believe that our analysis remains informative within the context of current clinical practice. For example, if a patient is starting care with a new physician, the physician needs to know the complete medical history of colonoscopy to inform clinical care (2), yet our results identified that patients may be unable to provide these necessary details. Therefore, a patient report of any previous polyp could be used as a marker to the clinician that additional information from previous colonoscopies needs to be located. Also, this study focused on determining characteristics associated with accurate self-report of polyp. Assessment of correlates of accurate adenoma self-report would also have been beneficial; however, the measures of accuracy for adenoma were so low that it was not useful to assess these associations.

Among a large population of participants in the PLCO SCU Trial, self-reported colonoscopy findings of a polyp were accurate but recall that an adenoma was present was poor. Identification of strategies to increase accurate self-report are needed, particularly those that can improve patient-provider communication and assist patient reporting of results to family members.

No potential conflicts of interest were disclosed.

The research presented in this paper is that of the authors and does not necessarily reflect the official position of the NCI or the NIH.

Conception and design: K.P. Wiseman, C.N. Klabunde, R.E. Schoen, V.P. Doria-Rose

Development of methodology: C.N. Klabunde, R.E. Schoen, V.P. Doria-Rose

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.E. Schoen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.P. Wiseman, C.N. Klabunde, D. Buckman, P. Wright, T.P. Hickey, R.E. Schoen, V.P. Doria-Rose

Writing, review, and/or revision of the manuscript: K.P. Wiseman, M.I. Silver, C.N. Klabunde, D. Buckman, R.E. Schoen, V.P. Doria-Rose

Funding to conduct this study was provided by intramural research funds.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.