Overtreatment in Nonmalignant Lesions Detected in a Colorectal Cancer Screening Program: A Cross- Sectional Analysis

Joaquin Cubiella (  joaquin.cubiella.fernandez@sergas.es ) Hospital Universitario de Ourense https://orcid.org/0000-0002-9994-4831 Antía González Direccíon Xeral de Saude Pública. Conselleria de Sanidade Raquel Almazán Dirección Xeral de SAude Pública. Conselleria de Sanidade Elena Rodríguez-Camacho Direccíon Xeral de Saúde Pública. Conselleria de Sanidade Raquel Zubizarreta Dirección Xeral de Saúde Pública. Conselleria de Sanidade Isabel Peña-Rey Lorenzo MD Dirección General de Epidemiología: Direccion General de Epidemiologia

In screening programs, the bene t gained by individuals should outweigh any harm; for example from overdiagnosis, overtreatment, false positives, false reassurance, uncertain ndings and complications. [5] Although performing diagnostic tests as well as related complications in CRC screening are well determined, [2,6] there is no such certainty regarding overdiagnosis and overtreatment. Overdiagnosis is de ned as the diagnosis of a medical condition or disease that would not cause symptoms or death during a patient's lifetime. In the case of CRC screening, treatment of overdiagnosed CRC and polyps should be called overtreatment. [7] Endoscopic resection of colorectal polyps is the key to reduce CRC incidence and mortality. [8] Although side effects are limited, mainly postpolypectomy syndrome, rectal bleeding and perforation, they account for most colonoscopy-related injury during CRC screening. [2] Endoscopic resection removes up to 90% of advanced complex polyps. [9] However, introduction of CRC screening programs has increased the number of colectomies due to benign polyps. In the US, up to 25% of colectomies were performed for nonmalignant polyps. [10] Related mortality and morbidity attains 0.8% and 25.3%, respectively. [11] However, there is little information regarding the incidence of surgery and associated risks in subjects with benign lesions detected in a CRC screening program [12] Thus, we decided to perform a retrospective crosssectional analysis in the rst round of the Galician (northwestern Spain) CRC screening program to determine the surgery rate in patients with non-malignant lesions detected on colonoscopy, surgeryrelated complications, motivation for the surgery and nally, the factors independently associated with surgery.

Study design:
We designed a cross-sectional retrospective multicenter study using the Galician (Northwestern Spain) CRC screening program database to identify patients. We included in this analysis all patients that underwent at least one colonoscopy in the rst round of the CRC screening program from its implementation (May 2013) until July 2019. We excluded patients with an invasive CRC as the nal diagnosis.

Description of the Galician CRC screening program:
Galician CRC mass screening is based on a biennial fecal immunochemical test (FIT) with a 20 µg/g of feces threshold. FIT is offered to subjects aged between 50 and 69 years. Until July 2019, 721,349 subjects were invited to participate in the screening program. The program was started in the Health Areas of Ferrol; Ourense; Pontevedra, Santiago and Lugo; A Coruña and Vigo in 2013, 2015, 2016 and 2017, respectively. The mass screening program is coordinated by the Public Health Department of the Conselleria de Sanidade. It is responsible for identi cation of subjects, invitation to participate, reception of FIT results, citation of patients with a positive result to perform a colonoscopy and nal assessment of endoscopic and histologic results. The main difference between the Galician program and other programs existing in other regions, is central coordination and management of patient follow-up depending on their risk according to EU guidelines for quality assurance on CRC screening recommendations. [13] Primary healthcare clinics are in charge of collecting FIT kits and assessment of subjects with a positive FIT prior to colonoscopy. The hospitals in each health area are responsible for FIT analysis, colonoscopies, histologic analysis and assessment and treatment of patients with a CRC. Finally, Coordination Unit personnel introduce data obtained from the differente sources in the screening program information system regarding CRC stage according to the AJCC classi cation, [14] the nal classi cation of patients with a positive result [13] in addition to several quality endoscopist indicators according to Spanish guidelines on quality in screening colonoscopy. [15] During the rst round, the participation rate and number of FIT positives in the rst round were 42% and 6.63%, respectively.

Baseline data:
From each patient, we collected the information available in the screening program database: sex, age, fecal hemoglobin concentration, performance status, associated medical illnesses graded according to the American Society of Anesthesiologists' Physical Status Classi cation (ASA grade), number of baseline colonoscopies, number of polyps, adenomas and size of the largest adenoma. Patients were classi ed as high risk (≥ 20 mm or ≥5 adenomas), intermediate risk (3-4 adenomas, 10-19 mm in size, villous histology or high grade dysplasia), low risk (1-2 tubular adenomas <10 mm in size) and no adenomas according to the European guidelines for quality assurance in CRC screening. [16] Data regarding the center and the endoscopist that performed the rst complete colonoscopy were collected. The adenoma detection rate (ADR) and number of colonoscopies performed during the rst round were calculated for each endoscopist in the rst round. Endoscopists were classi ed into quartiles according to their ADR and number of colonoscopies performed. Finally, hospitals were classi ed according to their complexity level (tertiary versus secondary).

Surgery:
We identi ed all the patients that required surgery after colonoscopy using the Spanish Health System′s Hospital Discharge Records Database (CMBD in Spanish) and the CRC screening program database. The CMBD includes information on hospital discharges using a list of clinical codes to establish the diagnosis that justi ed the admission The CMBD database receives noti cations from approximately 98% of Spanish public hospitals. [17] Mandatory health insurance covers an estimated 99.5% of the Spanish population, although subjects not covered by health insurance can still receive treatment in public hospitals. All subjects included in the Galician CRC screening programme are attended in the Galician Public health System. The International classi cation of diseases (ICD) codes used to identify colorectal surgeries were: ICD-9-MC 48.6 over the period 2013-2015 and ICD-10-ES ODT(C,E-N,P)(0,4)ZZ; ODB(C,E-N,P)(0,4)ZZ and ODBP7ZZ over the period 2016-2019. We subsequently manually searched the clinical records of identi ed patients to con rm that colorectal surgery was related to the screening colonoscopy. Moreover, we retrieved the following data: reason for surgery, type of surgery, length of hospital stay and complications either during hospitalization or after discharge. We searched the clinical information in IANUS, the Galician electronic health record system that covers both all the Galician hospitals and the primary healthcare centers. In hospital complications were classi ed according to the Clavien-Dindo classi cation. [18] We classi ed surgery complications as minor if they were grade I-II and major if they were grade III-V. If surgery was due to resection of colorectal lesions, we collected data regarding size, morphology according to the Paris classi cation, [19] location, endoscopic resection and histologic ndings in the endoscopic and surgical specimen. Based on endoscopic reports we calculated the Size, Morphology, Site and Access (SMSA) score and we classi ed lesions accordingly. [20] 5. Analysis: First, we descriptively analyzed the subjects included. We reported quantitative and qualitative variables as median and interquartile range (IQR), and total number and percentage, respectively. Thereafter, we calculated the surgery rate according to the different variables assessed. We performed a univariate analysis using the Chi-square test for qualitative variables and the Student's t test for quantitative variables to determine those related to surgery. Finally, we included statistically signi cant or clinically relevant variables in a multivariable analysis using logistic regression (forward conditional) to determine which variables were independently related to surgery. We performed a secondary analysis after excluding transanal surgery to determine the colectomy rate, related complications and independently associated factors. Associations were expressed as Odds Ratio (OR) with a 95% con dence interval (CI).
Statistical analyses were performed with IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY, USA: IBM Corp.

Ethics issues:
The local Institutional Review Board assessed and approved the study (code 2018/593). As long as the study was based on database operation, no informed consent was required. The information was accessed according to prevailing European and Spanish legislation.

Description of the sample
Between May 2013 and June 2019, a total of 16,720 subjects underwent at least one colonoscopy during the rst round of the Galician CRC screening program. We excluded 1013 subjects with CRC as the nal diagnosis from this analysis. Therefore, we included in the analysis 15,707 patients with non-malignant lesions detected in the rst round. After linking this data with the CMBD database, we identi ed 352 patients with any of the codes related to colorectal surgery. After verifying the clinical records, we con rmed that 162 underwent colorectal surgery related to participation in the screening program, four due to colonoscopy-related complications and 158 due to resection of colorectal lesions ( Figure 1 and Table 1). The surgery rate was as follows: global: 10.3‰ (95% CI 8.7-11.9), due to colonoscopy complications: 0.2‰ (95% CI 0.005-0.5) due to resection of colorectal lesions: 10.0‰ (95% CI 8.5-11.6).
In the seven hospitals taking part in the CRC screening program, the surgery rate ranged widely between 2.7‰ (95% CI 0.5-5.0) and 18.9‰ (95% CI 13.6-24.3). After excluding transanal surgeries (31) the colectomy rate was 8.3‰ (95% CI 6.9-9.8). The colectomy rate again ranged between 2.3‰ (95% CI 0.3-4.3) and 16.2‰ (95% CI 10.6-21.7). In Table 1, we outline the characteristics of the sample as well as the surgery rate according to dependent variables and in Supplementary Table 1 we show the same results  referred to colectomies. Seventy-one endoscopists from seven hospitals took part in the rst round of the CRC screening program.
The median number of colonoscopies performed was 278 (IQR 56-507) and median ADR was 65.3% (IQR 60.0%-70.08%). We classi ed endoscopists into quartiles according to number of colonoscopies performed and ADR. Finally, hospitals were classi ed into tertiary (three) and secondary (four).

Characteristics of the resected colorectal lesions
As shown in Table 3, most surgically resected lesions were either right-sided (49.6%) or located in the rectum (22.2%). Median endoscopic size was 35 mm and most lesions were either sessile, at or laterally spreading tumors. The lesions had a SMSA score above 12 in most cases (76.7%). An endoscopic resection was attempted in 23.5% of patients either in the work-up colonoscopy or in scheduled therapeutic colonoscopy. Median surgical size of the lesion was 25 mm and, as in the endoscopic histology, the most common histology was adenomatous (81.8%).

Factors associated with surgery.
During univariate analysis, several factors related to the patient (age), screening program (FIT result and number of baseline colonoscopies performed), characteristics of the lesions detected (number of polyps and adenomas, adenoma size and classi cation according to the European guidelines for CRC screening), endoscopist quality metrics (ADR and number of colonoscopies performed), and hospital complexity were signi cantly associated with the surgery rate as shown in Table 2. With respect to colectomy, we also identi ed several associated variables as shown in supplementary

Discussion
This is the rst study that reports the surgery rate in patients with nonmalignant lesions detected within a mass CRC screening program and the factors related to it. Most surgeries are related to resection of colorectal lesions and, exceptionally, to endoscopic complications. This information is extremely relevant to measure overtreatment risks in this setting. Fortunately, the surgery rate is low and the associated risk of mortality is as expected: one related death in 15,000 subjects. However, we have determined that not only factors associated with the patient and endoscopic ndings but also the endoscopist's performance measured with the ADR and hospital level of complexity are independently associated with surgery rate.
Our study has several strengths. The rst is related to its population-based perspective. We have collected data from the rst round of the Galician CRC screening program. During this initial round, FIT was offered to 721,349 subjects aged 50-69 years, colonoscopies were performed in seven hospitals and quality indicators of seventy-one endoscopists were collected in a centralized database. This database enabled us to calculate surgery rate according to the different variables available. Thus, we could accurately determine the risk of overtreatment in a mass screening program and side effects related to surgery. There is not much information available to compare our data. As an example, in a retrospective study performed within the scope of the national English Bowel Cancer Screening Program, surgery rate in large polyps (≥20 mm at or sessile) attained 21.7%. [21] Our data are not comparable as long as the European guidelines high risk group includes adenomas of any morphology ≥ 20 mm in size and/or ≥5 adenomas. In fact, most surgically resected lesions in our study were either sessile, at or laterally spreading tumors (93%) with a SMSA >12 in most cases. A French study evaluated the frequency and risk factors for the surgical resection of non-malignant polyps detected in a FIT based CRC screening program. [12] In this study, the surgery rate in patients with any polyp detected was 4.1% and was related with factors related to the size, location, histology, endoscopy center and the endoscopist. In our study, the surgery rate in the patients with at least one adenoma is clearly inferior.
We have detected an association between the ADR and surgery rate independent from the endoscopic ndings. ADR is the endoscopist's main quality indicator and has been associated with the risk of interval CRC, [22] CRC death, [23] detection of serrated polyps [24] and the adenoma detection during surveillance. [25] Although ADR is considered a surrogate for meticulous inspection of the colorectal mucosa, correlation with other important outcomes has never been found. In our case, we hypothesize that our ndings re ect an association between assessment of the mucosa and the endoscopists' resection skills. Out results con rm that endoscopists are a risk factor for surgery in patients with polyps detected in a screening program. [12] Nevertheless, we must draw attention to the high ADR of endoscopists taking part in the Galician screening program. Although an ADR above 45% is recommended in FIT-based screening programs, [26] in our case 75% of endoscopists attained a 60% ADR.
Our study has several limitations related to quality of the data collected in the CRC screening database. First, we used the CMBD to identify all the colorectal surgeries. Although we do not have information regarding the accuracy of the data obtained from the Spanish CMBD, an evaluation of the ICD-9-CM for CRC in an Italian administrative database showed a sensitivity ranging between 98 and 99%. [27] Unfortunately, information regarding location, morphology of the most advanced lesion, SMSA classi cation or the visual predicted histology of the lesions detected was not stored. We cannot provide information regarding on the visual suspicion of malignancy of the polyp that could explain some of the referrals to surgery. Additionally this information could explain one of the most striking ndings of our study. Although males have an increased risk of advanced neoplasia detection in CRC screening [28] and account for 75% of the high risk lesions detected, [29] in our study, females have an increased risk of surgery. The reason is unclear and we suggest it may be related to differences in the natural history of CRC. There is evidence that the serrated carcinogenic pathway, through hypermethylation and BRAF V600E mutation, at or sessile serrated lesions located proximally, [30,31] is more common among females and this could explain our ndings. In fact, as the study by Le Roy et al [12] shows, location is a risk factor for surgery in polyps detected in a screening program. Unfortunately, this information was unavailable for patients not requiring surgery.
Colorectal complications and mortality in our study are as expected. Data analyzed from a National Surgical Quality Improvement Program from 2011 to 2014, including 12,732 patients who underwent elective surgery for nonmalignant colorectal lesions, revealed a 0.7% 30-day mortality rate and 14% risk of major postoperative adverse events. [10] We also analyzed the long term complications that mainly impair the subject`s quality of life. [32] In contrast, endoscopic resection is more cost-effective, has few side effects, complications no greater than 1% to 2% and mortality below 1/10,000. [2] In large colorectal lesions, endoscopic resection-related mortality ranges between 0 and 0.08%. [33,34] Despite professional society guidelines and recommendations, [9,35] colectomies for benign colon lesions have increased in the last few years. In the US, surgery incidence for nonmalignant lesions has increased from 5.9/100,000 to 9.4/100,000 adults in 2000 and 2014, respectively. [10] Our study highlights the need for improved endoscopic resection techniques. First, endoscopists need to be trained speci cally in visual assessment of colorectal lesions and in resection techniques and their results should be continuously monitored. In this sense, we require contrasted quality indicators adapted to each screening scenario (FIT, colonoscopy). Quality indicators such as visual diagnostic yield, complete resection, complications, relapse and colectomy rates in large colorectal polyps should be monitored both per endoscopist and per endoscopy unit. However, complex endoscopic resection techniques such as submucosal dissection and endoscopic full thickness resection should be available and patients should be referred to centralized units where these techniques are performed on a regular basis. [35] Conclusions To conclude, the surgery rate in patients with nonmalignant lesions detected in a mass screening program is low and mainly associated with treatment of unresectable polyps. Although complications related to surgery are acceptable, this is an area both endoscopists and endoscopy units can improve upon. In order to reduce the number of subjects referred to surgery, we need to improve endoscopist resection skills and centralized units for complex techniques should be available. Finally, we require endoscopic resection quality indicators that enable us to continuously monitor endoscopic resection results.

Declarations
Ethics approval and consent to participate The local Institutional Review Board of Pontevedra-Vigo-Ourense, Spain assessed and approved the study (code 2018/593). As long as the study was based on database operation, no informed consent was required. The information was accessed according to prevailing European and Spanish legislation.

Consent for publication:
Not required.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests:
The authors have no con ict of interest to declare. Joaquín Cubiella has nancial support from Instituto de Salud Carlos III but has no nancial relationships with any organizations that might have an interest in work submitted in the previous ve years, and no other relationships or activities that could appear to have in uenced the submitted work.
Funding: This study was supported by grants from Academia Médico Quirúrgica of Ourense, Spain and the Instituto de Salud Carlos III through project PI17/00837 (Co-funded by European Regional Development Fund/European Social Fund "A way to make Europe"/"Investing in your future"). Ciberehd is funded by Instituto de Salud Carlos III. The funders had no role in the content of the manuscript nor in the decision to publish it.

Authors' contributions
The authors' contributions were as follows: JC, AG, RA, ERC, IPRL and RZ participated in the study design and search for nancial support; AG, RA, ERC, took part in operation of the CRC screening database; JC and AG participated in design of the study database; JC, took part in the clinical records search; JC performed the statistical analysis and wrote the rst draft. Finally, all the authors made comments on the article and approved the submitted version of the manuscript. All authors had full access to all the data (including statistical reports and tables) in the study and can take responsibility for the integrity of the data and accuracy of the data analysis. JC had full access to all of the data in the study and takes responsibility for the integrity of the data and accuracy of the data analysis. JC acts as guarantor of the article.