Age and sex differences in two types of colon polyps and what they mean for colorectal cancer screening

We aimed to provide contemporary estimates for SSL prevalence and identify patient and procedure characteristics associated with SSL detection.

Our study was a retrospective observational study of an outpatient colonoscopy population at a non-tertiary hospital. Patient and procedure characteristics were recorded from the clinical records. Study approval was obtained from the East Metropolitan Health Service Governance, Evidence, Knowledge, Outcomes Committee (activity approval 32886) and did not require written or informed consent because of the low-risk, retrospective, and non-interventional design of the study.

All colonoscopies performed in an outpatient day surgery unit between Jan 1 and Dec 31, 2019, were included in the study (n = 2097). Patient data were obtained from referral letters and clinical records. Data recorded included age, sex, body mass index (BMI), and colonoscopy indication. Patients undergoing surveillance for inflammatory bowel disease (IBD) were not excluded. Patient characteristics that were not consistently documented, such as ethnicity, smoking status, and alcohol consumption, were not included. Colonoscopy data obtained included endoscopist speciality, quality of bowel preparation, withdrawal time, colonoscopy findings including the size and location of colorectal polyps, the presence of mass lesions, and incidental findings such as diverticular disease and haemorrhoids.

Polyp histopathology reports were reviewed, and histologic characteristics including the presence of dysplasia were recorded. Polyp types of interest were SSLs, adenomas, HPs, and TSAs. Adenomas were subclassified as tubular, tubulovillous, or villous adenomas. Advanced adenomas were defined as adenomas with a size ≥ 10 mm, a prominent villous component, or high-grade dysplasia. Clinically significant serrated polyps (CSSPs) were defined as SSLs, TSAs, proximal colon HPs ≥ 5 mm, or HPs ≥ 10 mm anywhere in the colon [20]. Molecular pathology information was not available in histopathology reports. We defined proximal colorectal polyps as polyps in the caecum, ascending colon, hepatic flexure, and transverse colon, while distal colorectal polyps were at or distal to the splenic flexure (i.e. splenic flexure, descending, sigmoid colon, and rectum) as previously described [21]. Individual endoscopist adenoma detection rate (ADR) and SSL detection rates (SDRs) were calculated.

Data are summarised as median (interquartile range), means (standard deviation), or proportions. Univariate analyses were performed to examine associations between the presence of SSLs or adenomas and individual patient and procedure characteristics using Pearson Chi-square tests or Student’s t tests. Missing data were omitted from analyses. All p values were reported as two-sided and were interpreted at the 5% level of significance. For correlations between SSL detection and adenoma detection, we included data on endoscopists who performed more than 100 colonoscopies during the study period. Multivariable logistic regression analyses were conducted, including patient demographic and colonoscopy-related characteristics that were significantly associated with the polyp types in univariate analyses. HPs were not included in multivariable predictive models for SSLs and adenomas, as they were not considered clinically significant. The outcome variables are SSL, CSSP, and adenoma detection. Odds ratios (OR) and 95% confidence intervals (CI) for SSL, CSSP, and adenoma detection are reported. OR is presented as adjusted OR (aOR) in multivariable analysis. All statistical analyses were performed with the Statistical Packages for the Social Sciences (SPSS), version 26 (IBM Corp., Armonk, NY, USA).

Amongst the 2097 colonoscopies performed in the study, 1365 (65%) involved polypectomy. The median patient age was 54 (41–63) years, and 46.6% of patients were female. The mean BMI was 26.5 (5.5) kg/m². The mean age did not differ significantly, comparing males versus females (51.8 years for males, 52.0 years for females, p = 0.798), nor did mean BMI (26.4 kg/m² for males, 26.6 kg/m² for females, p = 0.406). Reasons for colonoscopy are summarised in Table 1. Some colonoscopies were performed for multiple indications.

Colonoscopies were performed by specialist gastroenterologists (64.9%) or general surgeons (35.1%), with nine colonoscopists performing more than 100 colonoscopies each during the study year. Colonoscopes used were the Olympus 180 series. Bowel preparation was good (reported as excellent or adequate) in 93% of colonoscopies based on colonoscopist impression. A total of 29.8% of patients had diverticular disease. Chromoendoscopy use was documented in < 1% of colonoscopies. Withdrawal times were only reported in 50% of colonoscopies, however, were consistently longer than 6 min where documented.

Fifteen pathologists were involved in polyp diagnosis in our study. SSL detection was 13.8% overall, with most SSLs located in the proximal colon (64.4% proximal vs. 35.6% distal, p < 0.001) and having no dysplasia (99%). Approximately 30.4% of SSLs without dysplasia were > 10 mm. The percentage of SSLs located in the proximal colon was not significantly different between sexes or between patients under and over 50 years (p > 0.05). Only one patient had a formal diagnosis of serrated polyposis, as consistent with recent WHO guidelines for gastrointestinal neoplasia [7]. The distribution of polyps is shown in Table 2. Most adenomas were tubular adenomas with low-grade dysplasia (90.3%), followed by tubulovillous adenomas with low-grade dysplasia (8.1%). Six patients had adenocarcinoma (0.29% of all colonoscopies). Nine out of 2097 patients had incomplete polyp data.

The standard quality indicator for colonoscopist performance in screening colonoscopy is the ADR, i.e., the proportion of colonoscopies performed in which at least one adenoma is detected and removed [41, 42]. The currently accepted benchmark ADR for endoscopists is ≥ 25% [43], which has been used by some studies to derive benchmarks for SSL detection. The correlation between ADR and SDR, however, remains uncertain, with certain studies showing moderate to high correlation (as reflected in our findings) and others suggesting otherwise [18, 44, 45]. The Cancer Council Australia (CCA) recommends a serrated lesion detection rate of > 10% [46]. Nevertheless, this benchmark includes HPs, which are more prevalent than SSLs but are not considered premalignant [4]. As for SSL-specific benchmarks, GESA proposes a 4% detection threshold for SSLs [19]. Such a rate, however, may be inadequate considering the rising prevalence rates of SSLs and our detection rate of 13.8%. Our findings, therefore, support the revision of benchmarks for SSL detection as additional data emerges. Previous studies indicate that the mortality benefit of colonoscopy screening has been limited to distal CRC [47–49]. This suggests that enhancing the detection of SSLs, which are mostly proximal in location [4], will significantly improve the effectiveness of colonoscopy screening.

The significance of our finding of higher rates of SSLs in younger females is uncertain, particularly given it is a novel finding that is not described extensively in the literature. A study from the USA provides reassurance that proximal SSLs have a relatively low likelihood of progressing to cancer in young people [50]. This is attributed to the fact that most young-onset CRC (onset < 50 years) is distal and exhibits different molecular characteristics than SSLs [51]. A more recent study by Hamoudah et al., however, suggests an increased risk of metachronous advanced colorectal neoplasia when small, serrated lesions and adenomas coexist [52]. This means that when accounting for other factors like synchronous adenoma detection and variable SSL prevalence and sex distribution, the association between SSLs and young-onset CRC becomes less clear. Our study found that SSLs frequently coexist with adenomas−a major precursor to young-onset CRC [50]. Furthermore, when considering Australian data, we find that 24% of cases of young-onset CRC diagnosed between 2001 and 2008 were proximal in location, meaning some could have been contributed to by SSLs or risk factors shared with adenomas [53]. While some studies have demonstrated a high prevalence of SSLs in a general colonoscopy population [54], few studies have investigated the significance of a potentially higher prevalence of SSLs in patients under 50 years. The fact that 16.8% of females under 50 years in our study had SSLs may warrant further consideration and examination of how such detection rates may correlate with young-onset CRC. Although a study by Lash et al. [27] demonstrates slow rates of SSL progression to dysplasia and carcinoma, another study by Oono et al. suggests that such progression may also occur rapidly [55]. Therefore, while most SSLs in our study occurred without dysplasia, the natural history of SSLs and the risk they pose to females under 50 years with potential susceptibility to SSLs remains unclear. Finally, although young-onset CRC tends to occur distally whereas SSLs tend to occur proximally, a considerable proportion of SSLs in our study were distal (> 30%), including in patients < 50 years. Further studies using prospective, multicentre, and contemporary histopathological data are required to elucidate the significance of these findings.

Our study has several strengths, including a large and reasonably well-characterised population and reliance on histopathologic descriptions of polyps. We observed a high photo-documented caecal intubation rate and reasonably high-quality bowel preparation. Additionally, our data involved specialist colonoscopists without outlier SDRs, which adds generalisability to our findings and reduces the influence of inter-endoscopist variability. We also provide data that is likely to reflect contemporary diagnostic criteria for serrated lesions. As for limitations, the retrospective design of our study means it is limited by some missing or inconsistently reported data (for example smoking and ethnicity) and unknown biases. It is also unclear what proportion of patients had a prior colonoscopy. However, less than 20% of colonoscopies were performed for surveillance (the majority, approximately 15%, being for previous polyp detection), suggesting that most procedures were not for polyp surveillance.

Data regarding IBD in our study was limited as IBD patients comprised < 1% of our patient cohort. Nevertheless, associations between IBD and SSLs remain unclear [56]. In terms of colonoscopy characteristics, withdrawal times were not consistently documented, which leads to uncertainty regarding adherence to the recommended > 6-min withdrawal times [57]. Similarly, chromoendoscopy use was poorly documented, suggesting the technique was rarely used in this cohort. Furthermore, bowel preparation was recorded based on colonoscopist impression (graded as excellent, adequate, fair, inadequate, or poor) rather than validated scoring systems such as the Boston Bowel Prep Score. [58] Regarding histopathology, there is likely to be variability in our study in terms of diagnostic criteria for SSLs given that multiple pathologists (15) were involved in polyp diagnosis. It was, therefore, difficult to ascertain what proportion of pathologists utilised the 5th edition versus the 4th edition of the WHO criteria for SSL diagnosis. However, all pathologists worked in a tertiary teaching hospital capacity. As newer criteria are more sensitive for SSL diagnosis, our findings likely underestimate rather than overestimate the true prevalence of SSLs. Addressing the above limitations requires further, high-quality, large prospective studies.

The authors declare no competing interests.