Baseline characteristics of participants recruited into the COLO-COHORT study: a UK-wide resource for colorectal cancer prevention

Abstract

Objective

To describe the design, recruitment, and baseline characteristics of the first phase of COLO-COHORT and to establish this population as a resource for colorectal cancer (CRC) prevention research.

Methods

COLO-COHORT is a prospective, multicentre observational study recruiting adults undergoing colonoscopy through screening and routine National Health Service referral pathways within a 32-site UK network. We present descriptive data from the first 4754 participants, recruited through 27 Phase 1 sites, including demographics, lifestyle factors, clinical characteristics, biospecimen availability and colonoscopy outcomes.

Results

Participants had a median age of 64 years (IQR 57–70), and 54.3% were male. Overweight (38.0%) and obesity (30.5%) were common, and cardiometabolic comorbidities were prevalent (11.8% with type II diabetes; 39.4% with hypertension). Colorectal neoplasia was identified at colonoscopy in over half of participants (56.1%), with histologically characterised data available on polyp type, size, morphology and anatomical location. Core demographic and clinical variables were available for >95% of participants, with colonoscopy findings recorded for all. Linked blood and stool microbiome biospecimens were available for approximately 80% and 40% of participants, respectively.

Conclusion

This paper establishes and characterises COLO-COHORT’s Phase 1 study cohort as a well-phenotyped resource, representing a population of individuals undergoing colonoscopy via UK screening and routine clinical pathways. This cohort will support and underpin future collaborative clinical, epidemiological and translational research in CRC prevention.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Large, well-characterised cohorts integrating colonoscopy findings with detailed lifestyle, clinical and biological data are needed to advance colorectal cancer (CRC) prevention research. COLO-COHORT represents one of the first large UK-wide prospective colonoscopy cohorts integrating detailed clinical, lifestyle and biological data.

WHAT THIS STUDY ADDS

• This study introduces COLO-COHORT, a large UK multicentre colonoscopy cohort with linked demographic, lifestyle, clinical, biospecimen and colonoscopy data. The paper demonstrates the scale, completeness and research potential of the cohort as a resource for CRC prevention research.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• COLO-COHORT provides a platform for collaborative clinical, epidemiological and translational research to improve CRC risk stratification, investigate disease pathways and support future prevention and screening strategies.

Introduction

Colorectal cancer (CRC) remains one of the most commonly diagnosed cancers globally and a significant cause of cancer-related mortality.¹ In the UK, around 44 000 cases are diagnosed every year, with CRC being diagnosed most commonly in people over 50 years of age.^{2 3} Prevention and early detection are critical in reducing CRC incidence and mortality. Pre-cancerous colorectal neoplasia, such as adenomatous and serrated polyps⁴—provide an opportunity to identify individuals at increased risk of CRC and to intervene before malignancy develops. Characterising the factors associated with all stages of colorectal neoplasia is therefore essential for informing prevention strategies, refining risk stratification in clinical and public health settings and optimising the use of endoscopy resources.

The COLO-COHORT study was established to: (1) develop a multi-factorial risk prediction model for colorectal neoplasia, (2) investigate the role of the gut microbiome in colorectal neoplasia and (3) establish a consent-for-contact (C4C) platform for future research.⁵ COLO-COHORT additionally provides a valuable resource for investigating a range of clinical and epidemiological questions in colorectal neoplasia, including the use of multiple faecal immunochemical tests (FITs) in symptomatic participants and investigating factors associated with sessile serrated lesions (SSLs). As one of its kind in the UK, and by collecting extensive data across multiple domains, the study also offers a unique opportunity to study individuals with and without colorectal neoplasia at colonoscopy, supporting future CRC prevention research.

This paper presents a descriptive overview of participants recruited during Phase 1 of COLO-COHORT, comprising the first 4754 individuals recruited to the study. This paper aims to document how the cohort was assembled, describe baseline participant characteristics and demonstrate the completeness and breadth of available data. By formally introducing this resource at an early recruitment milestone, this paper seeks to support transparency, facilitate future collaboration and provide a foundation for subsequent hypothesis-driven collaborative analyses as recruitment continues.

Methods

Study design and setting

COLO-COHORT is an ongoing multi-centre observational study conducted across 32 hospital sites in the UK. Adults aged ≥30 years scheduled to undergo colonoscopy are identified and recruited through National Health Service (NHS) research sites. Patients referred for colonoscopy through the Bowel Cancer Screening Programme (BCSPs), as well as routine clinical referral pathways (henceforth, non-BCSP)—including symptomatic referral, family history assessment and post-polypectomy or post-cancer surveillance—are eligible to participate. Recruitment to Phase 1 of the study, reported here, took place between 1 January 2020 and 31 July 2024 and represented a predefined recruitment milestone.

COLO-COHORT’s published protocol⁵ specifies recruitment targets of 10 000 participants for observational Group A—including 6000 in A1 and 4000 in A2—and 10 000 participants for the Group B consent-for-contact platform (COLO-SPEED), with recruitment planned over a 6-year period (2019–2025). Participants recruited to Group A may also participate in Group B; these targets therefore do not necessarily represent 20 000 unique individuals.

COLO-COHORT was designed to allow prospective follow-up through linkage to routinely collected NHS datasets. Ethical approval and participant consent were obtained to enable linkage to national cancer registration, hospital episode statistics and mortality records. While the current analysis focuses on cross-sectional baseline data from Phase 1 recruitment, planned data linkage will allow longitudinal follow-up for CRC incidence, related clinical outcomes, resource utilisation and mortality.

The recruitment process is summarised in figure 1. Further information on the study is available in the published protocol.⁵

Figure 1. Participant recruitment and assessment process for Phase 1 of the COLO-COHORT study. ^aExcept for two sites that completed the above on the day of the procedure. ^bExcept for one site that provided the IPAQ to participants following colonoscopy to return with the FFQ. FFQ, Food Frequency Questionnaire; IPAQ, International Physical Activity Questionnaire.

This study is reported in accordance with the STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) guidelines for cohort studies; the completed checklist is provided in the online supplemental material.

Study population and data collection

Participants completed baseline demographic and lifestyle assessments and had colonoscopy findings recorded. Depending on the clinical pathway and recruitment arm (see protocol paper),⁵ some participants also (optionally) provided blood samples and preprocedural FIT samples. Availability of specific data types, therefore, varied between participants. For this descriptive analysis, all participants were considered a single cohort, and no comparisons were made based on consent category or recruitment subgroup.

Data were captured using a centralised REDCap database⁶ and underwent routine regular quality assurance procedures coordinated by the study centre, including range and internal consistency checks.

Demographics and clinical characteristics

Demographic data were collected by site research nurses, typically via telephone on the day prior to colonoscopy; a small number of sites collected these data in person on the day of the procedure (figure 1). Information recorded included age, sex, ethnicity, educational attainment, employment status, medical history, medication use, family history of CRC and other cancers and smoking status. Smoking status was subsequently categorised using pack-year categories (never, <10, 10–19, 20–29 and ≥30 pack-years). Anthropometric measurements, including height, weight and waist circumference, were measured by trained staff on the day of colonoscopy using standardised procedures. Waist circumference was measured in a standing position at the midpoint between the lower rib and the iliac crest using a non-stretch tape measure. Body mass index (BMI) was then calculated as weight in kilograms divided by height in metres squared. An overview of the data collected, including data completeness by instrument/data category, is summarised in online supplemental table 1.

Lifestyle factors

Diet

Dietary intake was assessed using the EPIC Food Frequency Questionnaire (FFQ),⁷ which captured habitual intake of 130 food items over the preceding 12 months. Responses were converted into daily food and nutrient intakes using the Food Extraction Tool for Analysis, based on the UK Nutrient Databank.⁸

Collection of dietary data was linked to the study’s gut microbiome substudy so the FFQ was administered only to participants whose stool sample underwent microbiome analysis: this comprised all participants recruited through non-screening (symptomatic/other) pathways UK-wide, and participants recruited through the BCSP specifically in the North East of England, where screening FIT samples are additionally processed through the coordinating centre’s microbiome pipeline.⁵ BCSP participants recruited elsewhere had screening FIT samples processed through their local regional hub only and were therefore not included in the microbiome sub-study or sent an FFQ (figure 1).

Physical activity

Physical activity was assessed using validated questionnaires. Participants recruited at study outset completed the EPIC Physical Activity Questionnaire (n=270), while later participants completed the short form of the International Physical Activity Questionnaire (IPAQ)⁹ (n=2616); this change was made to reduce participant burden. Physical activity data were expressed as metabolic equivalent task hours per week, and values derived from both instruments were standardised to this common metric to enable descriptive summarisation across data collection instruments.

FIT

Quantitative FIT samples were obtained pre-procedurally from participants, with participants instructed to collect samples prior to bowel preparation (according to consent and availability criteria). FIT concentrations were recorded as micrograms of haemoglobin per gram of faeces (µg Hb/g), with values outside analyser detection limits assigned the corresponding reporting threshold for descriptive purposes. Participants recruited through the BCSP had FIT samples taken as part of routine screening, while those recruited through other clinical pathways submitted a FIT sample at recruitment (research FIT), regardless of prior testing. In English symptomatic services, colonoscopy is typically offered to individuals with a FIT result ≥10 µg Hb/g, although thresholds vary across UK nations and clinical referral pathways. In English BCSP, colonoscopy is offered to individuals with a FIT result ≥120 µg Hb/g (currently being rolled out at 80 µg Hb/g); in Scotland, screening is offered at 80 µg Hb/g. FIT data were linked to clinical and colonoscopy findings for descriptive purposes. Where patients had previously submitted FIT samples for clinical analysis purposes, those FIT results were also recorded (where available).

Gut microbiome

Stool samples submitted for FIT testing were also used for gut microbiome analysis in participants who provided appropriate consent and from sites where collection could be performed. Microbiome sequencing data were available for a subset of participants, with 1762 individuals undergoing sequencing, as previously described,¹⁰ with linked dietary and clinical data enabling future integrative analyses. Once the FIT sample was analysed, pseudonymised samples were transferred to the Central Biobank Facility at Newcastle University and stored at −80 °C on arrival. Detailed laboratory and bioinformatic methods for microbiome sample processing and analysis are reported separately.¹⁰ Gut microbiome data are available through the European Nucleotide Archive (ENA; www.ebi.ac.uk/ena/) under accession number PRJEB91658. Previous research by the team has demonstrated excellent agreement between microbiome profiles obtained from FIT samples and those from fresh stool analysis.^{11 12}

Ascertainment of colorectal neoplasia

Colonoscopy findings were recorded for all participants, including the presence or absence of polyps and CRC. For polyps, information was collected on anatomical location, size, morphology, retrieval status and for retrieved polyps on histological subtype and dysplasia status. Neoplasia details were determined based on histological findings from hospital pathology reporting systems, with colorectal cancer site and stage recorded where applicable.

Neoplasia outcomes identified at colonoscopy were categorised as agreed a priori based on histological findings, with case definitions summarised in online supplemental table 2. These categories were developed following multidisciplinary discussion within the research team and informed by UK post-polypectomy and post-CRC surveillance guidelines.¹³ Recognising that surveillance guidelines are influenced by both risk stratification and healthcare resource considerations at the patient level, neoplasia data were also coded flexibly to allow reclassification or regrouping of outcomes according to the specific aims of future analyses. A decision was made to include diminutive rectal hyperplastic polyps in the ‘no neoplasia’ group because of the lack of neoplastic potential, the current recommendation is not to remove them¹³ and consequently, documentation and removal of these polyps are variable. This overall approach supports descriptive reporting and enables analyses at both the participant and lesion level.

Colonoscopy procedure details

Colonoscopy procedural details recorded included use of sedation and antispasmodic agents, withdrawal time (in minutes), performance of rectal retroflexion, completion of colonoscopy, quality of bowel preparation and an anonymised identifier for the endoscopist performing the procedure.

Blood samples

Blood samples were collected for routine clinical assays, including a full blood count, liver function tests, aspartate aminotransferase, C-reactive protein, lipid profile, fasting glucose and HbA1c testing. If relevant blood test results from the previous 8 weeks were available in the patient’s records, repeat testing was not undertaken. A separate sample was taken for DNA extraction. All clinical blood tests—excluding those for DNA extraction—were labelled and processed according to local NHS Trust protocols in local laboratories. The sample for DNA was pseudonymised using the unique study ID and transferred to the Central Biobank Facility at Newcastle University. Samples were stored at −80°C in preparation for planned future DNA extraction.

Medications and supplement use

Information on medication use was collected at baseline as part of the clinical assessment. Participants were asked about use of selected medications relevant to CRC risk, including aspirin, statins, metformin and antibiotics. An open-ended field was also included to capture additional medications, from which use of proton pump inhibitors was identified. Information on calcium and vitamin D supplementation was also recorded.

Data quality assurance

Regular data checking and cleaning were conducted by the coordinating centre, with a different hospital site reviewed every 2 months. This entailed tracking down missing data, checking for implausible values, and following up with sites. In addition, approximately 5% of all paper Case Report Forms and FFQs were requested from hospital sites and cross-checked against REDCap entries to identify discrepancies and errors. This was carried out in two batches: the first (‘Dataset 1’) included participants who consented up to 30th June 2023 and showed a discrepancy rate of 0.9%; the second (‘Dataset 2’) included participants who consented between 1 July 2023 and 31 July 2024, with a discrepancy rate of 1.1%. The final dataset was exported to Stata (version 18) for further validation, outlier assessment and variable coding prior to analysis.

Statistical analysis

Participant characteristics were summarised using descriptive statistics, with the median and IQR reported for continuous variables and frequencies (with percentage values) for categorical variables. Descriptive summaries are presented overall and, where relevant, by recruitment pathway. Dietary intake, physical activity, anthropometric measures, smoking status and cardiometabolic comorbidities were summarised across colorectal lesion categories

For dietary variables, analyses were restricted to participants recorded as having returned an FFQ. Dietary data met the predefined inclusion criteria when no more than 10 FFQ food items were missing and total energy intake was plausible (women: 500–3500 kcal/day; men: 800–4000 kcal/day). Of 1802 participants recorded as having returned an FFQ, 1674 (92.9%) met these criteria and were included in dietary summaries. Dietary intake values were adjusted for total energy intake to account for potential misreporting.¹⁴

Patterns of missing data were examined descriptively, and analyses were conducted using available data for each variable, with exclusions applied only where specified.

Results

Participant characteristics

Between 1 January 2020 and 31 July 2024, 4764 participants were recruited into Phase 1 of the COLO-COHORT study. Ten participants subsequently withdrew, leaving 4754 participants included in this analysis. Of these, 49% were recruited through the BCSP and 51% through symptomatic or other referral pathways.

Smoking exposure varied across pack-year categories: 45.1% of participants were never smokers (0 pack-years), 17.7% were light smokers (<10 pack-years), 11.7% were moderate smokers (10–19 pack-years), 8.9% were heavy smokers (20–29 pack-years) and 15.5% were very heavy smokers (≥30 pack-years) (table 1).

Table 1. Baseline demographic and clinical characteristics of Phase 1 participants.

Risk factor	Total	BCSP	Non-BCSP
Age, years	64 (57–70)	64 (60–70)	62 (52–70)
Sex
Female, n (%)	2171 (45.7)	869 (37.6)	1302 (53.3)
Male, n (%)	2582 (54.3)	1443 (62.4)	1139 (46.6)
Education level, n (%)
Less than secondary/high school	1897 (39.9)	841 (36.4)	1056 (43.2)
Secondary/high school	969 (20.4)	437 (18.9)	532 (21.8)
University degree or equivalent professional qualification	1875 (39.4)	1024 (44.3)	851 (34.8)
Height, cm	1.70 (1.62–1.78)	1.73 (1.65–1.79)	1.69 (1.61–1.77)
Weight, kg	82.0 (70.0–95.0)	83 (72.0–95.3)	80.0 (68.3–93.4)
BMI, kg/m2	27.8 (24.8–31.8)	27.9 (25.0–31.6)	27.8 (24.5–32.1)
BMI classification, n (%)
Underweight (<18.5 kg/m2)	54 (1.14)	21 (0.91)	33 (1.35)
Normal weight (18.5–24.9 kg/m2)	1210 (25.5)	546 (23.6)	664 (27.2)
Overweight (25–29.9 kg/m2)	1805 (38)	944 (40.8)	861 (35.3)
Obese (≥30–39.9 kg/m2)	1451 (30.5)	690 (29.8)	761 (31.2)
Severe obesity (≥40 kg/m2)	196 (4.12)	91 (3.94)	105 (4.3)
Waist circumference, cm	99.5 (90-109)	101 (92–110)	98 (88–108)
Self-reported health conditions
History of diabetes, n (%)
Type I	38 (0.8)	22 (0.95)	16 (0.66)
Type II	561 (11.8)	237 (10.3)	324 (13.3)
No	4155 (87.4)	2053 (88.8)	2102 (86.1)
History of hypertension, n (%)
Yes	1872 (39.4)	997 (43.1)	875 (35.8)
No	2879 (60.6)	1313 (56.8)	1566 (64.1)
History of bowel polyps, n (%)
Yes	904 (19)	293 (12.7)	611 (25)
No	3844 (80.9)	2018 (87.3)	1826 (74.8)
History of colorectal cancer, n (%)
Yes	178 (3.74)	15 (0.65)	163 (6.68)
No	4576 (96.3)	2297 (99.4)	2279 (93.3)
History of other cancers, n (%)
Yes	632 (13.3)	313 (13.5)	319 (13.1)
No	4116 (86.6)	1996 (86.3)	2120 (86.8)
History of angina, n (%)
Yes	417 (8.77)	190 (8.22)	227 (9.3)
No	4332 (91.1)	2121 (91.7)	2211 (90.5)
Medication history
Aspirin, n (%)	519 (10.9)	250 (10.8)	269 (11.0)
Statin, n (%)	1809 (38.1)	929 (40.2)	880 (36.0)
Metformin, n (%)	402 (8.5)	176 (7.6)	226 (9.3)
Antibiotics in last 6 months, n (%)	909 (19.1)	351 (15.2)	558 (22.9)
Proton-pump inhibitor, n (%)	1773 (37.3)	707 (30.6)	1066 (43.7)
Lifestyle factors
Calcium/vitamin D supplement use, n (%)
Yes	1395 (29.3)	652 (28.2)	743 (30.4)
No	3349 (70.4)	1654 (71.5)	1695 (69.4)
Physical activity, tertiles n (%)
Low	979 (20.6)	384 (16.6)	595 (24.4)
Medium	970 (20.4)	497 (21.5)	473 (19.4)
High	937 (19.7)	506 (21.9)	431 (17.6)
Smoking, n (%)
Never smoker (0 pack-years)	2143 (45.1)	1035 (44.8)	1108 (45.4)
Light smoker (<10 pack-years)	841 (17.7)	393 (17.0)	448 (18.3)
Moderate smoker (10–19 pack-years)	556 (11.7)	252 (10.9)	304 (12.4)
Heavy smoker (20–29 pack-years)	425 (8.9)	213 (9.2)	212 (8.7)
Very heavy smoker (≥30 pack-years)	739 (15.5)	392 (17.0)	347 (14.2)
Alcohol, g/d	3.4 (0–10.2)	7.0 (1.6–12.5)	3.3 (0–9.9)

^*Median and IQR or n (%); Percentages are calculated using the full column denominator. Missing or unknown values are not shown; consequently, percentages may not sum to 100%.

BCSP, Bowel Cancer Screening Programme.

Participant characteristics varied by referral pathway. Individuals recruited through BCSP routes were slightly older on average, more likely to be male, and less likely to report current smoking or obesity compared with those recruited through non-screening pathways. Detailed characteristics by recruitment pathway are presented in table 1.

Phase 1 participants were recruited from 27 hospital sites: 26 in England (n=4636) and one in Scotland (n=118), within the wider 32-site COLO-COHORT network. Demographic and clinical characteristics varied across geographical areas of recruitment, including age distribution, sex, educational attainment, BMI and prevalence of cardiometabolic comorbidities. These differences illustrate the geographic and sociodemographic diversity of the cohort. Regional characteristics are summarised in online supplemental table 3, and Phase 1 recruitment numbers by site are presented in online supplemental table 4.

FIT measurements

FIT results were available for the majority of participants (97.2%), defined as having at least one recorded FIT result, including routine screening FIT results for BCSP participants and research FIT results for those recruited through non-screening pathways. Among non-BCSP participants with research FIT measurements, 1670 (72.3%) had values below 10 µg Hb/g, the threshold used to guide symptomatic referral in England. These research measurements should not be interpreted as the clinical FIT results used to determine referral. Screening programme FIT levels were above the applicable programme threshold of 120 µg Hb/g in England or 80 µg Hb/g in Scotland. The distribution of research FIT values is shown in online supplemental figure 1, illustrating the number of values and range of quantitative FIT data available for future analyses.

Blood biochemistry

Blood biochemistry results were available for 97% of participants, with data completeness summaries presented by recruitment pathway in online supplemental table 1. In addition, a total of 10 532 whole blood samples were collected and stored for future genomic analyses.

Colonoscopy findings and colorectal neoplasia

Colonoscopy findings were recorded for all participants. A total of 8149 polyps were identified across the cohort, of which 8036 (98.6%) had available histological confirmation. Polyp characteristics, including histological subtype, size, morphology, anatomical location and dysplasia status, are summarised in table 2.

Table 2. Characteristics of polyps detected at colonoscopy among Phase 1 participants (n=4754).

Polyp characteristic	N (%)*
Total number of polyps	8036
Histological type	8036
Tubular adenoma	4417 (55.0)
Hyperplastic	1200 (14.9)
Tubulovillous adenoma	721 (8.97)
Sessile serrated lesion	659 (8.20)
Non-polypoid tissue	252 (3.14)
Inflammatory	108 (1.34)
Cancer	68 (0.846)
Villous adenoma	56 (0.697)
Other	48 (0.597)
Mixed polyp	38 (0.473)
Traditional serrated adenoma	15 (0.187)
Missing*	454 (5.65)
Size (mm)†	8036
<5	4246 (52.8)
5–9	2234 (27.8)
≥10	1452 (18.1)
Missing	104 (1.29)
Dysplasia present	8036
High-grade dysplasia	141 (1.75)
Low-grade dysplasia	5036 (62.7)
None	727 (9.05)
Missing/unknown	2132 (26.5)
Morphology	8036
Sessile (Is)	5036 (62.7)
Pedunculated (Ip)	918 (11.4)
Subpedunculated (Isp)	462 (5.75)
Flat elevated (0-IIa)	1334 (16.6)
Flat elevated and depressed (0-IIa/c)	29 (0.361)
Flat (0-IIb)	35 (0.436)
Flat depressed and elevated (0-IIc/a)	3 (0.037)
Depressed (0-IIc)	3 (0.037)
Missing	216 (2.69)
Location	8036
Proximal (3893)
Caecum	886 (11.0)
Ascending colon	1344 (16.7)
Hepatic flexure	286 (3.56)
Transverse colon	1377 (17.1)
Distal (4135)
Splenic flexure	312 (3.88)
Descending colon	666 (8.29)
Sigmoid	2206 (27.5)
Rectum	951 (11.8)
Missing	8 (0.10)

^*Histology reports not updated or returned by hospitals.

^†Based on the endoscopy report.

At participant level, 2666 (56.1%) of 4754 participants had colorectal neoplasia identified at colonoscopy (table 3). This included 2117 participants (44.5%) with at least one adenoma, 888 (41.9%) with at least one advanced adenoma and 250 (5.3%) with CRC. At least one serrated lesion (defined as hyperplastic polyps, SSLs and traditional serrated adenomas) was identified in 923 participants (19.4%), including 193 (20.9%) with at least one advanced serrated lesion (serrated polyp >10 mm or serrated polyp with dysplasia) and 402 (8.5%) with one or more SSLs.

Table 3. Participant-level colorectal neoplasia outcomes identified at colonoscopy among Phase 1 participants (n=4754).

		BCSP	Non-BCSP
Outcome category	n (%)*	n (%)*	n (%)*
No neoplasia	2088 (43.9)	591 (25.6)	1497 (61.3)
Low risk	1526 (32.1)	872 (37.7)	654 (26.8)
High risk	890 (18.7)	664 (28.7)	226 (9.3)
Cancer†	250 (5.3)	185 (8.0)	65 (2.7)
Any adenoma	2117 (44.5)	1451 (68.5)	666 (31.5)
Non-advanced adenoma	1229 (58.1)	786 (54.2)	443 (66.5)
Advanced adenoma	888 (41.9)	665 (45.8)	223 (33.5)
Any serrated lesion	923 (19.4)	609 (66.0)	314 (34.0)
Non-advanced serrated lesion	730 (79.1)	486 (79.8)	244 (77.7)
Advanced serrated lesion	193 (20.9)	123 (20.2)	70 (22.3)
Sessile serrated lesion (SSL)	402 (8.5)	296 (48.6)	106 (33.8)
Overlapping profiles
Adenoma+SSL	259 (5.4)	208 (9.0)	51 (2.1)
Cancer+SSL	22 (0.5)	21 (0.9)	1 (0.1)
Cancer+adenoma	115 (2.4)	96 (4.2)	19 (0.6)
Cancer+serrated polyps	44 (0.9)	39 (1.7)	5 (0.2)
Serrated polyps+adenoma	579 (12.2%)	435 (18.8%)	144 (5.9%)
Any large non-pedunculated colorectal polyps	171 (3.6)	114 (4.9)	57 (2.3)

^*Each outcome is reported independently. Participants may meet criteria for more than one category (eg, adenoma and SSL); therefore, totals exceed 100%.

^†Out of the 250 cancers, 105 (42.0%) were found in the distal colon, 81 (32.4%) in the rectum and 70 (28.0%) in the proximal colon; six participants had cancers recorded at more than one anatomical location.

BCSP, Bowel Cancer Screening Programme.

A total of 579 participants (12.2%) had both adenomatous and serrated lesions, 259 (5.4%) had both adenoma and SSL, and smaller proportions had cancer in combination with adenoma or serrated lesions (table 3).

Lesion distribution by demographic and clinical characteristics

Colorectal lesion distributions varied by demographic and clinical characteristics. Adenomas were more frequently observed in men, who accounted for 65.2% of participants with adenomas. The majority of colorectal lesions occurred in participants aged over 50 years, comprising more than 90% of cases across lesion categories (table 4).

Table 4. Demographic, lifestyle, clinical and procedural characteristics according to colorectal lesion status.

	Total	No neoplasia	Any adenoma	Non-advanced adenoma	Advanced adenoma	Any serrated lesion	Sessile serrated lesion only	Cancer only
Demographic characteristics
Age
<50 years	498 (10.5)	376 (18)	72 (3.4)	45 (3.66)	27 (3.04)	57 (6.18)	8 (9.3)	4 (4.35)
>50 years	4256 (89.5)	1712 (82)	2045 (96.6)	1184 (96.3)	861 (97)	866 (93.8)	78 (90.7)	88 (95.7)
Sex
Female	2171 (45.7)	1163 (55.7)	737 (34.8)	468 (38.1)	269 (30.3)	357 (38.7)	44 (51.2)	47 (51.1)
Male	2582 (54.3)	924 (44.3)	1380 (65.2)	761 (61.9)	619 (69.7)	566 (61.3)	42 (48.8)	45 (48.9)
Lifestyle factors
Diet
Total energy intake, Kcal/d	1637 (1283–2046)	1656 (1294–2050)	1625 (1268–2034)	1627 (1263–2023)	1621 (1286–2036)	1612 (1224–1970)	1615 (1324–1805)	1729 (1261–2401)
Fruit intake, g/d	142.6 (61.9–252.3)	151.4 (66.9–254.7)	125.8 (55.6–244)	129.8 (61.9–242.7)	119.2 (47.3–245.2)	127.6 (51–241.3)	226.5 (100–339)	148.3 (62.6–231.9)
Vegetable intake, g/d	224.1 (139.5–311.9)	225.2 (139.5–314.4)	219 (143.2–306.9)	223.9 (146.7–314.3)	213.6 (139–302.1)	225.8 (137.1–309.4)	232.6 (160.8–288.6)	215.2 (121.8–321.5)
Red meat, g/d	20.3 (7.5–42.6)	19.3 (6.9–40.3)	24.2 (13.4–49)	24.8 (13.4–49)	22.2 (13.9–53.3)	21.7 (13.4–46)	19.3 (13.4–49)	17.1 (8.1–53.4)
Processed meat, g/d	19.6 (8.4–35.7)	17.4 (7.5–34.3)	21 (10.4–38.5)	20.7 (8.4–37.8)	23 (11.3–41.3)	20.7 (10.4–35.7)	22 (8.4–34.2)	18.7 (13.6–47.2)
Dietary calcium, mg/d	849.8 (647.3–1079.9)	858.9 (665.1–1104.9)	825.9 (629.4–1039.9)	834.6 (645.5–1054)	806.4 (614.1–1022.2)	804.2 (638.4–1042.4)	921.4 (658.1–1076.7)	937.9 (647.8–1116.3)
Alcohol intake, g/d	3.4 (0–10.2)	2.7 (0–9.1)	5.4 (0.5–11.9)	5 (0–11.2)	6 (1–13.6)	5 (0–10.3)	5 (0–10.3)	5.3 (0–21)
Physical activity
Inactive	909 (34.7)	419 (36.2)	379 (32.8)	238 (35.6)	141 (29)	175 (33.2)	12 (25.5)	19 (38)
Moderately active	1205 (46.1)	536 (46.3)	529 (45.8)	283 (42.4)	246 (50.5)	241 (45.7)	23 (48.9)	24 (48)
Highly active	502 (19.2)	202 (17.5)	247 (21.4)	147 (22)	100 (20.5)	111 (21.1)	12 (25.5)	7 (14)
BMI
Underweight   (<18.5 kg/m2)	54 (1.14)	31 (1.49)	18 (0.85)	11 (0.9)	7 (0.79)	5 (0.54)	0 (0.0)	0 (0.0)
Normal weight   (18.5–24.9 kg/m2)	1210 (25.5)	611 (29.3)	474 (22.4)	283 (23)	191 (21.5)	192 (20.8)	19 (22.1)	29 (31.5)
Overweight   (25–29.9 kg/m2)	1805 (38)	715 (34.2)	877 (41.4)	506 (41.2)	371 (41.8)	374 (40.5)	32 (37.2)	35 (38.0)
Obese   (≥30–39.9 kg/m2)	1451 (30.5)	622 (29.8)	652 (30.8)	378 (30.8)	274 (30.9)	303 (32.8)	31 (36)	27 (29.3)
Smoking
Never smoker (0 pack-years)	2143 (45.1)	1035 (49.6)	873 (41.2)	524 (42.6)	349 (39.3)	336 (36.4)	34 (39.5)	43 (46.7)
Light smoker (<10 pack-years)	841 (17.7)	387 (18.5)	357 (16.9)	208 (16.9)	149 (16.8)	140 (15.2)	16 (18.6)	23 (25.0)
Moderate smoker (10–19 pack-years)	556 (11.7)	230 (11.0)	265 (12.5)	157 (12.8)	108 (12.2)	134 (14.5)	15 (17.4)	6 (6.5)
Heavy smoker (20–29 pack-years)	425 (8.9)	162 (7.8)	200 (9.4)	111 (9.0)	89 (10.0)	115 (12.5)	7 (8.1)	8 (8.7)
Very heavy smoker (≥30 pack-years)	739 (15.5)	254 (12.2)	398 (18.8)	214 (17.4)	184 (20.7)	190 (20.6)	12 (14.0)	12 (13.0)
Clinical characteristics
History of diabetes
No	4155 (87.4)	1840 (88.1)	1846 (87.2)	1072 (87.2)	774 (87.2)	799 (86.6)	76 (88.4)	75 (81.5)
Yes	599 (12.6)	248 (11.9)	271 (12.8)	157 (12.8)	114 (12.8)	124 (13.4)	10 (11.6)	17 (18.5)
History of hypertension
No	2879 (60.6)	1378 (66)	1181 (55.8)	686 (55.8)	495 (55.7)	525 (56.9)	61 (70.9)	45 (48.9)
Yes	1872 (39.4)	708 (33.9)	936 (44.2)	543 (44.2)	393 (44.3)	398 (43.1)	25 (29.1)	47 (51.1)
Procedural/lesion characteristics
Indication route
BCSPs	2312 (48.6)	591 (28.3)	1451 (68.5)	786 (64)	665 (74.9)	609 (66)	51 (59.3)	55 (59.8)
Non-BCSP	2442 (51.4)	1497 (71.7)	666 (31.5)	443 (36)	223 (25.1)	314 (34)	35 (40.7)	37 (40.2)
Polyp location
Proximal	679 (14.3)	N/A	467 (22.1)	360 (29.3)	107 (12.1)	188 (20.4)	57 (66.3)	3 (3.26)
Distal	988 (20.8)	N/A	686 (32.4)	367 (29.9)	319 (35.9)	236 (25.6)	24 (27.9)	15 (16.3)
Both	1058 (22.3)	N/A	963 (45.5)	501 (40.8)	462 (52)	499 (54.1)	5 (5.81)	0 (0.0)

^*Median and IQR; percentages may not total 100% due to missing data, which are not shown in the table.

BCSP, Bowel Cancer Screening Programme; BMI, body mass index; N/A, not applicable.

Lifestyle and clinical data across lesion categories

Overall distributions varied across lesion groups, reflecting the breadth of lifestyle data available within the study population. For illustrative purposes, median processed meat intake ranged from approximately 19 to 23 g/day across lesion categories, compared with 17.4 g/day among participants with no neoplasia detected (table 4, online supplemental table 6). Detailed descriptive summaries of dietary intake, physical activity, anthropometric measures and smoking status are provided in online supplemental table 1 to characterise the range and completeness of available data.

Anatomical distribution varied by lesion type, with serrated lesions more commonly located in the proximal colon (66.3%) and advanced adenomas more frequently detected in the distal colon (35.9%). Lesion prevalence also differed by referral pathway, with advanced neoplasia more commonly identified among participants recruited through the BCSPs than through non-screening routes (table 4, online supplemental table 6).

Discussion

This paper describes the baseline characteristics of the first 4754 participants recruited to Phase 1 of the COLO-COHORT study, a large UK-wide, multi-centre observational cohort designed to support CRC prevention research. By characterising participants recruited through both screening and other routine clinical pathways (mainly symptomatic referral) and by documenting the availability and completeness of lifestyle, clinical, procedural and biological data, this analysis establishes COLO-COHORT as a well-phenotyped data resource to support future clinical, epidemiological and translational research.

A key strength of COLO-COHORT is the availability of histologically confirmed colonoscopy outcomes, with detailed, comprehensive information on polyp type, size, morphology, dysplasia status and anatomical location. This level of clinical detail enables flexible categorisation of neoplasia outcomes and supports future analyses focused on specific lesion subtypes, including advanced adenomas and SSLs. Notably, the number of serrated lesions identified in COLO-COHORT represents one of the largest series reported within a colonoscopy research cohort to date. Given the growing recognition of the malignant potential of serrated lesions and their contribution to colorectal carcinogenesis, such scale and phenotypic detail are particularly valuable.¹⁵ Despite increasing awareness, the aetiology and modifiable risk factors associated with serrated lesions remain incompletely understood, highlighting the importance of a well-phenotyped cohort such as COLO-COHORT for advancing colorectal prevention research.^{16 17}

In this study population, more than half of the participants undergoing colonoscopy were found to have colorectal neoplasia, including adenomas, serrated lesions and CRC. The prevalence and anatomical distribution of lesions observed are consistent with patterns reported in colonoscopy-based populations, including a higher frequency of adenomas in men and a predominance of serrated lesions in the proximal colon.^16–21 These findings provide reassurance regarding the completeness and validity of lesion-level ascertainment and reflect the heterogeneous indications for colonoscopy across screening and non-screening referral pathways, including the increasing recognition and detection of serrated lesions as part of contemporary CRC prevention practice.

Participants in COLO-COHORT exhibited a high prevalence of established CRC risk factors, including overweight and obesity, cardiometabolic comorbidities and prior smoking. Lifestyle and behavioural data were comprehensively captured, including dietary intake, physical activity, anthropometric measures and smoking history. While this analysis was not designed to assess associations between lifestyle factors and colorectal neoplasia, the descriptive summaries presented illustrate the breadth and depth of exposure data available for future hypothesis-driven analyses. The inclusion of validated instruments for diet and physical activity assessment aligns with approaches commonly used in nutritional and lifestyle epidemiology and supports comparability with existing literature.^{9 22 23} The distribution of overweight and obesity, cardiometabolic comorbidities and smoking history observed in COLO-COHORT is consistent with established CRC risk factor profiles reported in the UK and international literature and with patterns commonly observed among individuals undergoing colonoscopy and CRC screening.^{23 24} Differences in lifestyle profiles between screening and non-screening populations have been reported previously and may reflect broader patterns of health-seeking behaviour and screening participation.²⁵ Together, these characteristics highlight the relevance of the population for future analyses examining the contribution of modifiable lifestyle factors to colorectal neoplasia risk.

COLO-COHORT also benefits from linkage to pre-procedural FIT results and from the availability of gut microbiome sequencing data in a substantial subset of participants. The use of FIT-derived stool samples for microbiome analysis has been previously validated,¹¹ enabling large-scale investigation of microbial composition and function in relation to colonoscopy-confirmed outcomes. These features position COLO-COHORT to address emerging research questions at the interface of lifestyle, host genomics, the gut microbiome and colorectal neoplasia, including the evaluation of strategies involving multiple FITs, an area of active investigation within CRC screening and symptomatic referral pathways.^{26 27} In particular, the availability of detailed colonoscopy outcomes, FIT results and linked lifestyle and biological data provides an opportunity to build on existing work, such as COLO-FIT, a UK-based initiative evaluating FIT-based triage algorithms for symptomatic referral pathways,^{28 29} and to explore whether additional factors may further inform risk stratification and referral algorithms used in clinical practice.

Few international cohorts combine colonoscopy-confirmed outcomes with detailed lifestyle and gut microbiome data at comparable scale. CRCbiome³⁰ in Norway is nested within a national FIT-based screening trial and combines lifestyle and gut microbiome data with screening outcomes, while the German BLITZ study³¹ recruits participants undergoing screening colonoscopy and collects questionnaire, blood and stool data to investigate colorectal neoplasia risk and non-invasive detection markers. Both cohorts are restricted to screening populations, whose underlying risk profiles and routes to colonoscopy are likely to differ from those of COLO-COHORT participants recruited through non-BCSP clinical pathways. COLO-COHORT’s inclusion of participants from both screening and non-BCSP pathways therefore provides a complementary resource spanning a broader range of colonoscopy indications.

Several limitations should be considered when interpreting the characteristics of the COLO-COHORT population. The cohort was predominantly White British, and a relatively high proportion of participants reported university-level education or equivalent qualifications. For example, recent census data indicate that approximately one-quarter to one-third of adults in comparable age groups in England and Scotland hold degree-level qualifications, suggesting that educational attainment in COLO-COHORT is higher than that of the general population.^{32 33} While recruitment spanned multiple sites across England and Scotland, demographic characteristics vary regionally, and therefore the distributions observed in COLO-COHORT should be interpreted in the context of a colonoscopy-attending population drawn from specific geographic areas rather than as representative of the UK population overall. Available national datasets, including colonoscopy audit data and the National Endoscopy Database,^{34 35} further indicate that individuals undergoing colonoscopy differ from the wider population with respect to sociodemographic characteristics. The patterns seen in COLO-COHORT may also reflect differential participation in colonoscopy-based studies and research more generally by socioeconomic and ethnic characteristics.^{36 37} Limited ethnic diversity in study populations remains a recognised challenge in UK public health and clinical research and has been identified as a priority area for improvement by major funders.³⁸ These considerations should be borne in mind when interpreting future analyses derived from this cohort.

Beyond ethnic and educational composition, a more fundamental concern (which applies to many, if not most, research studies) relates to participation bias arising from the nature of consent to COLO-COHORT itself. Enrolment, for some, requires completion of a 130-item FFQ and a physical activity questionnaire, provision of stool and blood samples and consent to linkage with national NHS datasets—a substantially greater commitment than is required simply to attend a scheduled colonoscopy. Individuals who agree to this level of participation are likely to differ systematically from those who decline, not only in ethnicity and educational attainment but in unmeasured characteristics including health literacy, engagement with healthcare services and broader health-seeking behaviour. Because these characteristics may themselves be correlated with the lifestyle and clinical risk factors that COLO-COHORT is designed to study, the educational attainment gradient observed here (for example) is likely to be a partial marker of a broader participation effect rather than its full extent. This has direct implications for one of COLO-COHORT’s stated aims—development of a multifactorial risk prediction model for colorectal neoplasia. Any such model derived from this cohort should be understood as describing risk relationships among people who both attend colonoscopy and consent to intensive research participation, and its generalisability to the wider colonoscopy-attending population, and more so to the general population, will require formal external validation before clinical application.

A related but distinct limitation concerns the population from which COLO-COHORT is drawn. By design, participants are recruited only after they have already been selected by the healthcare system for colonoscopy, either through engaging with screening and having a positive test or clinical referral following symptoms. This excludes individuals who do not participate in screening, do not present with or report symptoms, or are not referred for investigation—groups that are arguably of greatest relevance to a population-level CRC prevention strategy, since they are the least likely to be reached by existing pathways. COLO-COHORT is therefore well placed to address questions that (for example) compare individuals with and without colorectal neoplasia among those who reach colonoscopy, for example, identifying clinical, lifestyle or biological features that distinguish adenoma-bearing from adenoma-free colonoscopy patients—but this is conceptually distinct from identifying risk factors for colorectal neoplasia at the population level. Researchers using COLO-COHORT data should therefore distinguish between these two classes of question: those concerning risk stratification and phenotyping within a colonoscopy-attending population, for which the cohort is well suited, and those concerning population-level prevention or screening uptake, which would require linkage to or comparison with non-attending or non-referred populations beyond the scope of the current resource.

A further consideration relates to the interpretation of research FIT results among participants recruited through non-BCSP pathways. Among those with a research FIT result, 1670 (72.3%) had values below 10 µg Hb/g. As described in the Methods, research FIT was collected at recruitment regardless of previous clinical FIT testing and was not necessarily the test used to determine referral. Furthermore, the non-BCSP group included symptomatic, family-history, post-polypectomy surveillance and post-cancer surveillance referrals, to which the 10 µg Hb/g symptomatic threshold does not apply uniformly. Low research FIT values are therefore expected and do not imply inappropriate referral. As 99.22% of non-BCSP participants were recruited after 1 June 2020, the predominance of values below 10 µg Hb/g cannot principally be attributed to recruitment before formalisation of FIT-based symptomatic triage. Future analyses should distinguish research FIT from clinical referral FIT and account for referral indication when comparing screening and non-screening pathways.

Furthermore, lifestyle exposures were assessed at baseline and relied on self-reported questionnaires, which are subject to errors in recall and, hence, misclassification.³⁹ However, the dietary and physical activity instruments used have demonstrated acceptable reproducibility and validity in population-based studies.^{7 9} Physical activity data were collected using two validated instruments over the course of recruitment, and although harmonised for descriptive purposes, direct comparability between instruments is limited. Future analyses using physical activity as a variable of interest should treat instrument type as a potential source of measurement heterogeneity, for example, by including it as a covariate or conducting instrument-stratified sensitivity analyses, rather than assuming full equivalence between the two harmonised measures.

Additionally, some lesion histology data were unavailable due to non-retrieval or deferred management; however, these instances were infrequent and reflect real-world clinical practice.

The availability of dietary data reflected both the protocol-defined scope of FFQ administration and subsequent questionnaire return and quality assessment. FFQs were linked to the microbiome sub-study and were therefore not intended for all Phase 1 participants. Among participants recorded as having returned an FFQ, 1674 of 1802 (92.9%) met the dietary analysis criteria; this included 193 of 215 BCSP participants (89.8%) and 1481 of 1587 non-BCSP participants (93.3%). Thus, only 128 returned FFQs (7.1%) were excluded for missing items or implausible energy intake. Researchers planning nutritional epidemiology analyses should report the dietary analytic subsample clearly and consider its concentration in the North East of England and in non-screening participants, which may introduce selection effects into diet-focused analyses.

The principal sources of potential bias, their likely effects on different research objectives, proposed mitigation strategies and opportunities for external validation are summarised in online supplemental table 5.

Despite these limitations, COLO-COHORT represents a uniquely comprehensive nationwide resource for CRC prevention research. Ongoing and planned analyses will focus on risk prediction modelling, microbiome–neoplasia relationships, serrated lesion biology and longitudinal outcomes, building on the descriptive characterisation presented here as recruitment and follow-up progress. More broadly, the national scale of COLO-COHORT, inclusion of participants from both screening and routine clinical pathways, high-quality colonoscopy outcomes and extensive lifestyle and biological data provide a strong foundation for future collaborative clinical, epidemiological and translational research to advance CRC prevention.

Data availability statement

Data are available upon reasonable request.