Nonadherence to Referral Colonoscopy After Positive Fecal Immunochemical Test Results Increases the Risk of Distal Colorectal Cancer Mortality

The fecal immunochemical test (FIT) with diagnostic colonoscopy is the recommended modality for population-based colorectal cancer (CRC) screening, but the adherence rate of colonoscopy remains suboptimal. Furthermore, it is demonstrated that individuals with a positive FIT result have a high detection rate of colorectal neoplasia at colonoscopy.1Zorzi M. et al.Gut. 2018; 67: 2124-2130Crossref PubMed Scopus (54) Google Scholar However, evidence on the long-term risk of CRC incidence and mortality among FIT-positive individuals who do not undergo a colonoscopy is limited, especially for young adults aged <50 years. Additionally, whether the long-term risk varies with anatomic site is unclear. To fill in the evidence gap, we aimed to investigate whether nonadherence to colonoscopy in individuals with a positive FIT result was associated with long-term risk of site-specific CRC incidence and mortality. We conducted a large cohort study under the framework of a population-based CRC screening program in China, covering 595,180 individuals aged 40–74 years since 2007. Participants who had a positive FIT result and were given a referral for colonoscopy were included in the study, including 42,353 adherers and 32,070 nonadherers of colonoscopy (Supplementary Figure 1). We calculated the hazard ratio (HR) of CRC incidence and mortality using Cox proportional and Fine-Gray regression models, respectively. See the Supplementary Methods for details. Baseline characteristics of the 2 groups are shown in Supplementary Table 1. The 10-year cumulative incidence and mortality of CRC were 12.84 and 2.48 per 1000 cases among adherers compared to 17.15 and 5.63 per 1000 cases among nonadherers, respectively (Figure 1). The risk of CRC (HR, 1.25; 95% confidence interval [CI], 1.09–1.43) and CRC-related death (HR, 1.85; 95% CI, 1.36–2.50) were higher among nonadherers vs adherers (Supplementary Table 2). When stratifying by tumor location, nonadherers had increased risk of distal CRC (HR, 1.36; 95% CI, 1.15–1.60) and its mortality (HR, 1.98; 95% CI; 1.38–2.85) but not proximal CRC (HR, 1.11; 95% CI, 0.86–1.44) and its mortality (HR, 1.35; 95% CI, 0.74–2.46). In age-stratified analysis, for those aged 40–49 years, nonadherers had a 2.01-fold increased risk of CRC (95% CI, 1.33–3.04), which was more remarkable than that in individuals aged 50–64 years (HR, 1.34; 95% CI, 1.10–1.62) and 65–74 years (HR, 1.06; 95% CI, 0.85–1.31), with a similar trend being observed in CRC mortality. In sex-stratified analysis, the results for CRC mortality showed no significant heterogeneity between men (HR, 1.93; 95% CI, 1.34–2.79) and women (HR, 1.91; 95% CI, 1.11–3.28). When doing sensitivity analysis by additional adjustment of diabetes and smoking among individuals with a first positive FIT result occurring during the second round of screening (since 2012), the results were basically consistent with those observed in the main analyses (Supplementary Table 2). This large population-based cohort study, with a follow-up period of up to 16 years, suggests increased risks of CRC and related death among FIT-positive individuals who did not undergo colonoscopy. Our findings are consistent with the results of 2 previous studies, which also demonstrated an increased risk of CRC-related death among nonadherers of colonoscopy.2Lee Y.C. et al.J Natl Cancer Inst. 2017; 109: djw269Crossref PubMed Scopus (40) Google Scholar,3Zorzi M. et al.Gut. 2021; 71: 561-567Crossref PubMed Scopus (29) Google Scholar The higher risk of dying from CRC in colonoscopy nonadherers can be attributed to both an increased likelihood of late-stage CRC diagnosed among nonadherers and to the early detection and removal of precancerous lesions facilitated by colonoscopy in adherers. Our study further supports the notion that the effectiveness of colonoscopy among FIT-positive individuals in reducing the incidence and mortality of distal CRC is more remarkable than that of proximal CRC.2Lee Y.C. et al.J Natl Cancer Inst. 2017; 109: djw269Crossref PubMed Scopus (40) Google Scholar The disparity of site-specific effectiveness may be related to the low sensitivity of colonoscopy in detecting proximal lesions. It has been shown that serrated lesions, particularly sessile serrated lesions, which are widely considered to be precursor lesions of CRC, are mostly located in proximal colon and are hard to be detected and completely removed by colonoscopy.4Meester R.G.S. et al.Gastroenterology. 2020; 159: 105-118.e25Abstract Full Text Full Text PDF PubMed Scopus (37) Google Scholar Therefore, certain approaches should be considered to improve the accuracy of colonoscopy for detecting proximal lesions and reduce the disparity of site-specific effectiveness in reducing CRC incidence and mortality, such as establishing a benchmark for sessile serrated lesion detection rate and developing novel techniques to improve the detection rate of serrated lesions.5Keswani R.N. et al.Gastroenterology. 2021; 161: 701-711Abstract Full Text Full Text PDF PubMed Scopus (45) Google Scholar We found that an elevated risk of CRC in nonadherers vs adherers were more remarkable for individuals aged <50 years than older adults. The underlying reason for the result was unclear. It may be related to both higher odds of false-positive FIT results and higher susceptibility to proximal CRC and serrated lesions for the old adults than for younger individuals.6Lash R.H. et al.J Clin Pathol. 2010; 63: 681-686Crossref PubMed Scopus (249) Google Scholar,7Wong M.C. et al.Gastrointest Endosc. 2015; 81: 596-607Abstract Full Text Full Text PDF PubMed Google Scholar Additionally, Jung et al8Jung Y.S. et al.Gastrointest Endosc. 2017; 86: 892-899Abstract Full Text Full Text PDF PubMed Scopus (14) Google Scholar found that FIT-positive individuals aged 35–49 years had a higher prevalence of advanced colorectal neoplasia than the 50–59-year adults at general risk. These findings indicate that young adults who had a positive FIT result and underwent a subsequent timely colonoscopy might reduce their long-term risk of CRC. Our study has several strengths, including the large screening-naive population, linkage to a cancer registry with high-quality and comprehensive coverage of cancer and death registration, and detailed subgroup analyses by age and tumor site. However, primary limitations also exist. First, unmeasured confounders may still exist, although we performed sensitivity analyses to additionally adjust for lifestyle factors, and the confounding effect of access to care was minimal because of equal access to free colonoscopy for all individuals and a standardized health care system involved in the screening program. Further studies are warranted to verify the findings by addressing more potential confounding factors. Second, the rate of colonoscopy performed outside the designated hospitals should be considered. A small follow-up survey of this study showed that less than 5% individuals underwent a paid colonoscopy at hospitals outside the screening program. Third, immortal time bias in the CRC incidence analysis may be introduced. However, it is important to note that the immortal person-years accounted for only approximately 1.8% of the total observation time in our study, thereby minimizing its impact on the results. In conclusion, individuals with a positive FIT result but who did not adhere to the referral colonoscopy had higher risk of CRC incidence and mortality. Moreover, the risk elevation was more pronounced among younger adults. The disparity of the site-specific effectiveness of the current screening strategy should be noted, and future efforts are needed to improve the benefits of colonoscopy screening for proximal CRC. The Colorectal Cancer Screening Cohort Research Group includes Chengcheng Liu,1 Yunfeng Zhu,2 Qilong Li,3 Mingyang Song,4,5 Jinhuan Yang,3 Tian Jin,1 Tianyi Ling,1 Yanqin Huang,6 and Mingjuan Jin7; from the 1Department of Colorectal Surgery and Oncology (Key Laboratory of Cancer Prevention and Intervention, China National Ministry of Education, Key Laboratory of Molecular Biology in Medical Sciences, Zhejiang Province, China), and The Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China; 2Haining Cancer Prevention and Treatment Research Institute, Chinese Medicine Hospital of Haining, Haining, Zhejiang, China; 3Jiashan Institute of Cancer Prevention and Control, Jiashan, Zhejiang, China; 4Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, Massachusetts; 5Department of Nutrition, Harvard T. H. Chan School of Public Health, Boston, Massachusetts; 6Cancer Institute, Key Laboratory of Cancer Prevention and Intervention, Ministry of Education, The Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China; and 7Department of Epidemiology and Biostatistics, Cancer Institute of the Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China. We would like to acknowledge all staff involved in the colorectal cancer screening program in Haining County and Jiashan County as well as all participants who contributed data to this study. Yingshuang Zhu, PhD (Conceptualization: Lead; Formal analysis: Lead; Funding acquisition: Supporting; Methodology: Lead; Visualization: Lead; Writing – original draft: Lead). Xue Li, PhD (Conceptualization: Equal; Funding acquisition: Supporting; Methodology: Equal; Writing – review & editing: Lead). Yeting Hu, PhD (Data curation: Lead; Formal analysis: Supporting; Project administration: Supporting; Writing – review & editing: Supporting). Kun Chen, PhD (Data curation: Equal; Project administration: Supporting; Writing – review & editing: Supporting). Shu Zheng, PhD (Conceptualization: Supporting; Project administration: Lead; Writing – review & editing: Supporting). Kefeng Ding, PhD (Conceptualization: Equal; Funding acquisition: Lead; Project administration: Equal; Supervision: Lead; Writing – review & editing: Equal). Supplementary Table 1Baseline Characteristics of Participants in Adherers and Nonadherers of ColonoscopyVariablesAdherersNonadherersParticipants, n42,35332,070Age, y, n (%) 40–4911,050 (26.1)8527 (26.6) 50–6423,357 (55.1)15,646 (48.8) 65–747946 (18.8)7897 (24.6) Median (IQR)56 (49–63)57 (49–64)Sex, n (%) Male21,160 (50.0)16,216 (50.6) Female21,193 (50.0)15,854 (49.4)Screening site, n (%) Jiashan15,937 (37.6)7939 (24.8) Haining26,416 (62.4)24,131 (75.2)Calendar year of recruitment, n (%) 2007–20106760 (16.0)3268 (10.2) 2011–201312,381 (29.2)6383 (19.9) 2014–20169632 (22.7)8947 (27.9) 2017–201913,580 (32.1)13,472 (42.0)Family history of CRC, n (%) Yes1010 (2.4)373 (1.2) No38,938 (91.9)30,537 (95.2) Missing2405 (5.7)1160 (3.6)Positive FIT results, n (%) 1 FIT+ from 1-sample test5277 (12.5)6245 (19.5) 1 FIT+ from 2-sample test29,191 (68.9)21,545 (67.2) 2 FIT+s from 2-sample test7885 (18.6)4280 (13.3)Degree of FIT positivity, n (%) Strong positive29,964 (70.7)14,918 (46.5) Weak positive12,389 (29.3)17,152 (53.5)Smoking, n (%) Yes10,290 (24.3)9476 (29.5) No25,025 (59.1)16,682 (52.0) Missing7038 (16.6)5914 (18.4)Diabetes, n (%) Yes1783 (4.2)1279 (4.0) No33,422 (78.9)23,498 (73.3) Missing7148 (16.9)7295 (22.7)CRC cases, n511393 Proximal colon, n (%)149 (29.0)100 (25.3) Distal colon or rectum, n (%)329 (64.1)277 (69.9) Unknown location, n (%)35 (6.8)19 (4.8)CRC deaths, n8196 Proximal colon, n (%)26 (32.1)21 (21.9) Distal colon or rectum, n (%)54 (66.7)70 (72.9) Unknown location, n (%)1 (1.2)5 (5.2)IQR, interquartile range. Open table in a new tab Supplementary Table 2Risk of Incidence and Mortality of CRC Among Nonadherers vs Adherers of ColonoscopyVariablesAdjusted 10-year cumulative incidence/mortality rate (95% CI)aAdjusted by age at the positive FIT result (continuous variable), sex, year of screening, family history of CRC, screening site, number of positive FIT results, and degree of FIT positivity.Crude HR (95% CI)P valueAdjusted HR (95% CI)AdherersNonadherersModel 1aAdjusted by age at the positive FIT result (continuous variable), sex, year of screening, family history of CRC, screening site, number of positive FIT results, and degree of FIT positivity.P valueModel 2bAdditionally adjusted by smoking and diabetes among individuals with the first positive FIT result occurring during the second round of screening (since 2012).P value CRC incidenceCRC12.2 (11.1–13.3)16.6 (14.9–18.4)1.12 (0.98–1.28).0921.25 (1.09–1.43).0011.26 (1.07–1.49).006Proximal CRC3.70 (3.05–4.35)4.40 (3.47–5.33)1.00 (0.77–1.29).9901.11 (0.86–1.44).4231.17 (0.84–1.63).351Distal CRC7.98 (7.04–8.92)11.9 (10.4–13.5)1.23 (1.05–1.44).0121.36 (1.15–1.60)<.0011.38 (1.13–1.68).001Age, y40–493.89 (2.70–5.08)8.71 (6.12–11.3)1.56 (1.05–2.34).0302.01 (1.33–3.04).0011.78 (1.05–3.00).03150–6411.5 (10.0–12.9)16.6 (14.1–19.1)1.07 (0.88–1.28).5101.34 (1.10–1.62).0031.27 (1.01–1.61).04465–7424.3 (20.5–28.1)27.9 (23.3–32.5)0.96 (0.78–1.18).6891.06 (0.85–1.31).6251.22 (0.94–1.60).138SexMale14.4 (12.6–16.2)20.9 (18.1–23.7)1.20 (1.01–1.42).0391.42 (1.20–1.70)<.0011.43 (1.16–1.76).001Female10.1 (8.65–11.5)12.4 (10.3–14.5)1.00 (0.81–1.24).9791.20 (0.97–1.49).0911.00 (0.75–1.32).990CRC mortality CRC2.56 (1.91–3.21)5.31 (4.03–6.59)1.92 (1.42–2.59)<.0011.85 (1.36–2.50)<.0011.97 (1.31–2.97).001Proximal CRC0.77 (0.42–1.12)1.13 (0.57–1.69)1.33 (0.74–2.42).3411.35 (0.74–2.46).3281.75 (0.62–4.95).292Distal CRC1.87 (1.27–2.47)4.21 (2.96–5.46)2.09 (1.46–3.00)<.0011.98 (1.38–2.85)<.0011.99 (1.27–3.13).003 Age, y40–490.67 (0.01–1.33)2.49 (0.22–4.76)2.45 (0.79–7.61).1223.13 (0.95–10.3).0602.87 (0.50–16.5).23650–641.68 (1.09–2.27)3.86 (2.45–5.27)1.67 (1.04–2.68).0352.12 (1.32–3.38).0021.84 (0.97–3.47).06165–747.27 (4.44–10.1)13.2 (8.51–17.8)1.65 (1.09–2.49).0191.60 (1.05–2.42).0282.04 (1.16–3.60).013 SexMale3.96 (2.70–5.22)8.18 (5.70–10.7)1.87 (1.30–2.69).0011.93 (1.34–2.79).0012.17 (1.35–3.49).001Female1.28 (0.73–1.83)2.66 (1.59–3.73)1.94 (1.14–3.31).0151.91 (1.11–3.28).0191.69 (0.74–3.87).214NOTE. The bold values signify P < .05.a Adjusted by age at the positive FIT result (continuous variable), sex, year of screening, family history of CRC, screening site, number of positive FIT results, and degree of FIT positivity.b Additionally adjusted by smoking and diabetes among individuals with the first positive FIT result occurring during the second round of screening (since 2012). Open table in a new tab IQR, interquartile range. NOTE. The bold values signify P < .05. This retrospective cohort study was conducted as part of an ongoing nationwide pilot study in a population-based CRC screening program, which was initially launched in Jiashan County and Haining County, Zhejiang Province, in January 2007. A FIT-based screening was implemented, and the details of this program have been described in our previous studies.1Zorzi M. et al.Gut. 2018; 67: 2124-2130Crossref PubMed Scopus (54) Google Scholar,2Lee Y.C. et al.J Natl Cancer Inst. 2017; 109: djw269Crossref PubMed Scopus (40) Google Scholar Briefly, permanent residents in Jiashan and Haining aged 40–74 years without severe hypertension, heart diseases, bleeding tendency, mental disabilities, or pregnancy were invited to participate in the screening program. The participants were interviewed by trained interviewers with a baseline questionnaire and were asked to provide 2 fecal samples with a 1-week interval for 2 FITs at each round of screening. Among those who returned fecal samples, the majority of participants (84.6%) provided 2 samples as requested. The FIT was conducted using a qualitative kit (Abon Biopharm Co, Ltd) with a detection threshold of 100 ng hemoglobin/mL buffer (20 μg hemoglobin/g feces). Participants with at least 1 positive FIT were defined as FIT-positive individuals and were contacted via telephone and advised to undergo colonoscopy examination in designated hospitals. Both the FITs and colonoscopy examination were free of charge. The participants in this study were individuals who participated in CRC screening in Jiashan and Haining from January 1, 2007, to August 31, 2019. Two rounds of screening had been conducted by that time. Individuals with a positive FIT in at least 1 round of screening were included. For those who had positive FITs in both rounds of screening, the first positive FIT would be used. We excluded individuals with a prior colorectal cancer/adenoma/polyp diagnosis or a history of colonoscopy. Furthermore, individuals who exhibited clinical symptoms such as diarrhea; constipation; mucous, bloody stool; abdominal pain;, or changes in bowel habits were excluded. For individuals undergoing a colonoscopy, those with an incomplete colonoscopy (eg, poor bowel preparation or did not reach the cecum), or a colonoscopy delayed by more than 12 months at the designated hospitals, or a colonoscopy with a missing check date were excluded. In addition, individuals who died within 1 year after the positive FIT result were also excluded. In this screening program, of those who underwent a referral colonoscopy, 99.6% of participants had their colonoscopy within 12 months of the positive FIT result. In this study, individuals who received a colonoscopy within 12 months after a positive FIT result were defined as the colonoscopy adherer group, those who were failed to undergo a colonoscopy were defined as the nonadherer group. Baseline characteristics of study participants were collected, including age, sex, family history of CRC, calendar year of screening, FIT results, and clinical symptoms. Starting from the second round of screening (since 2012), additional information on comorbidities (ie, diabetes) and lifestyle factors (ie, smoking) was also collected. Smoking status was defined as consuming at least 1 cigarette per day for more than 1 year or consuming more than 300 cigarettes within 3 months. The colonoscopy quality measures (eg, bowel preparation and cecum intubation) and clinical characteristics of polyps were documented in the medical records of the CRC screening program. The outcomes of this study were CRC incidence and mortality. Information on cancer diagnosis and death was obtained by linking the unique personal identification number with the Haining and Jiashan Cancer and Death Registry Systems. CRC was defined as International Classification of Diseases, 10th Revision codes C18–C20. Cancer located in the cecum to the transverse colon (C18.0–C18.4) was defined as proximal CRC, and cancer located in the splenic flexure to the sigmoid colon (C18.5–C18.8) as well as the rectum (C19, C20) was defined as distal CRC. Cancer with International Classification of Diseases, 10th Revision code C18.9 was defined as unknown location of CRC. All CRC cases and deaths underwent medical record review. For CRC incidence, the follow-up started from the positive FIT result to the date of CRC diagnosis, death, or the end of follow-up (December 31, 2022), whichever came first. Individuals who died before a CRC was diagnosed or remained alive with no CRC diagnosed at the end of follow-up were referred as censoring data. For CRC mortality, the follow-up started from 1 year after the positive FIT result (time 0) to the date of death or the end of follow-up, whichever came first. Individuals who died from non-CRC causes or were alive at the end of follow-up were referred as censoring data. Descriptive statistics were used to describe the distribution of baseline characteristics in the adherer and nonadherer groups. Continuous variables are presented as median and interquartile range. The cumulative CRC incidence and mortality in the 2 groups were estimated using the Kaplan-Meier method. For CRC mortality, we used the Fine-Gray regression models, considering CRC death as an event of interest and deaths from other causes as competing risks. In the context of CRC incidence analysis, where the influence of death as a competing risk was minimal, we chose the Cox proportional regression models to calculate HRs. Colonoscopy adherers were regarded as the reference group. Age, sex, year of screening, family history of CRC, screening site, number of positive FIT results, and degree of FIT positivity (based on the color of the FIT strip) were included as covariates for adjustment in models. The missing values were treated as a separate group for each covariate. The adjusted cumulative incidence and mortality rates were also calculated by including the aforementioned covariates and using an SAS macro (%ADJSURV).3Zhang X. et al.Comput Methods Programs Biomed. 2007; 88: 95-101Crossref PubMed Scopus (246) Google Scholar Furthermore, we conducted subgroup analyses to assess the effectiveness of colonoscopy in reducing the CRC incidence and mortality in the proximal and distal sites. Additionally, we performed stratification analyses to estimate the association of colonoscopy nonadherence with the risk of CRC incidence and mortality in different age groups (40–49 years, 50–64 years, and 65–74 years) and by sex. Sensitivity analysis was performed by additionally adjusting for diabetes (yes/no) and smoking (yes/no) among individuals with the first positive FIT result occurring during the second round of screening (since 2012). SAS version 9.4 software (SAS 9.4, SAS Institute Inc, Cary, NC) was used for all analyses, and a 2-sided P value of less than .05 was considered to be statistically significant.