Modifiable Risk Factors for Esophageal Cancer in Endoscopic Screening Population: A Modeling Study.
Qian Zhang,Fan Wang,Hao Feng,Jie Xing,Shengtao Zhu,Hao Zhang,Yang Li,Wenqiang Wei,Shutian Zhang
DOI: https://doi.org/10.1097/cm9.0000000000002878
IF: 6.133
2023-01-01
Chinese Medical Journal
Abstract:To the Editor: Esophageal cancer, one of the most common cancer types in China, with an estimated 346,633 new cases and 323,600 deaths in 2022, is becoming an increasingly serious clinical and public health problem.[1] The successful promotion of the self-management strategy has indicated that lifestyle modifications can be valuable in the primary prevention of cancer development. Adopting a healthy lifestyle has become a novel strategy for primary prevention and risk reduction in high-risk areas. Previous epidemiological studies have identified several lifestyle-related risk factors for esophageal cancer, including smoking and diet.[2] Each factor can typically explain a modest proportion of cancer risk. However, when combined, these known risk factors may substantially affect the risk of esophageal cancer. Nevertheless, some risk factors for esophageal cancer are non-modifiable, including age, low socioeconomic status, and family history. Whether and how these non-modifiable risk factors affect primary cancer prevention by intervening with modifiable risk factors remain unclear. This study aimed to determine how the combined non-modifiable and modifiable risk factors modified the esophageal cancer risk and evaluate whether the risk of esophageal cancer can be decreased by optimizing the modifiable risk factors with different risk levels based on non-modifiable factors. We developed an absolute risk model that has a flexible time duration for risk prediction and is compatible with variable modifications, thereby distinguishing the variables into modifiable and non-modifiable risk components. In addition, we determined the rank of the main independent modifiable risk factors and the number of preventable cases when lifestyle-related risk factors were modified to the lowest risk level along with non-modifiable risk factors. Using data from 7342 individuals aged 40–69 years who underwent endoscopic screening and questionnaire interviews and participated in two esophageal cancer screening cohorts [Supplementary Figure 1, https://links.lww.com/CM9/B760],[3,4] a population-based absolute risk model combining the effects of lifestyle modifications and non-modifiable risk factors interventions was established with estimated odds ratios (ORs), population incidence, and mortality from the National Central Cancer Registry of China and the Third National Retrospective Sampling Survey of Death Causes. The risk stratification and population attributable risk owing to non-modifiable and modifiable risk factors were evaluated via this model. Sensitivity analyses were performed to investigate the robustness. Participants in the two screening cohorts were followed up by the second household follow-up survey or the household registration system at local police stations. Follow-up data, including date of esophageal cancer onset and disease coding using the International Classification of Diseases codes (ICD-10), were recorded. The follow-up study was approved by the Ethics Committee of Beijing Friendship Hospital (approval No. 2017-P2-142-01) and the Institutional Review Board of the Cancer Foundation of China (approval date: 2008-10-06). Information regarding the demographic characteristics (age, sex, family income, family history of cancer, and education level) and modifiable lifestyle-related risk factors (water source, smoking status, alcohol consumption, and dietary habits, including eating foods from the following categories: pickled food; fried food; onions and garlic; meat, eggs, and milk; and fruits) was collected using baseline questionnaires in both cohorts. Family income was adjusted by the changes in disposable income per capita. Using the randomization stratified by disease status, the total analysis set was then split into a random training set (N = 3671) and a testing set (N = 3671). A multivariate logistic regression model with all demographic variables and modifiable lifestyle-related risk factors was applied in the training set to estimate ORs and the corresponding log OR parameter β, which was used to predict the absolute risks [Supplementary Table 1, https://links.lww.com/CM9/B760]. The absolute risk of developing esophageal cancer for a subject at a certain age was calculated according to the methodology in the "iCARE" package of R software Version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria).[5] Data on age-specific disease incidence and competing hazards of mortality were collected from the Report on the Third National Retrospective Sampling Survey of Death Causes and the National Central Cancer Registry of China, respectively. To predict the cumulative risk of esophageal cancer, the population was stratified into six subgroups (G1–G6) based on the quartile of the total risk scores for absolute risks (subgroups 1–6: 0–19%, 20–39%, 40–59%, 60–79%, 80–89%, and ≥90% quartiles of risk scores). Similar to the total absolute risk scores, to assess the risk modification of modifiable risk factors, the population was also classified into six strata (S1–S6) based on non-modifiable risk factor scores (strata 1–6: 0–19%, 20–39%, 40–59%, 60–79%, 80–89%, and ≥90% quartiles of non-modifiable risk scores). In these six non-modifiable risk strata, the modifiable lifestyle-related risk factors were set as observed risks (no modification of these lifestyle factors), the lowest risks (all lifestyle factors modified to the lowest levels; e.g., smoking cessation; not eating pickled, fried food, and fruits; eating onions and garlic; eating meat, eggs, and milk), and significant modifiable risk factors (the only modified significant factors were smoking cessation and eating onions and garlic). The PAR percentage (PAR%) was predicted to evaluate the percentage of esophageal cancer cases that could be prevented by optimizing the modifiable risk factors with the population at different risk levels stratified based on non-modifiable factors. Detailed information is provided in the Supplementary Material, https://links.lww.com/CM9/B760. Our established model revealed that age, water source, and family history were significant non-modifiable factors. Meanwhile, smoking (OR = 2.32, 95% confidence interval [CI]: 1.04–5.18, P = 0.040) and eating onions and garlic (OR = 0.52, 95% CI: 0.31–0.89, P = 0.016) were identified as the main significant modifiable lifestyle-related risk factors. The AUC to predict esophageal cancer was 0.86 (95% CI: 0.85–0.88) in the training set and 0.87 (95% CI: 0.85–0.88) in the testing set [Supplemental Figure 2, https://links.lww.com/CM9/B760]. The absolute risks with combined non-modifiable and modifiable risk factors for developing esophageal cancer over the next 50 years for a disease-free 30-year-old individual from the underlying population was predicted to be 1.63% on average, ranging from 0.10% to 9.14% for the lowest and highest subgroups [Figure 1A]. For individuals in the highest strata of risk owing to non-modifiable factors (S5 and S6), the median absolute risks explained by the modifiable risk factors (smoking status; alcohol consumption; pickled food; fried food; onions and garlic; meat, eggs, and milk; and fruits) were much higher than the average risks [Figure 1B]. Those who modified these risk factors to the healthy levels could significantly decrease risks to the average level in the general population [Figures 1C–D]. The combined modification of the identified significant risk factors resulted in the prevention of 69.20% of esophageal cancer cases [Supplementary Table 2, https://links.lww.com/CM9/B760]. Modification of a single risk factor was less effective than the combined modification method but could still help explain the remarkable proportion of preventable cancers.Figure 1: The absolute cumulative risk derived from the starting age of 30 years to a certain age for esophageal cancer (A) and the cumulative risk of esophageal cancer by risk strata via lifestyle modification ([B] without applying any modification; [C] with modifications to all lifestyle items; and [D] with modifications only to smoking and onions and garlic intake). Note: G1: subgroup 1 of total absolute risk scores (risks ranged from 0 to 19%); G2: subgroup 2 of total absolute risk scores (risks ranged from 20% to 39%); G3: subgroup 3 of total absolute risk scores (risks ranged from 40% to 59%); G4: subgroup 4 of total absolute risk scores (risks ranged from 60% to 79%); G5: subgroup 5 of total absolute risk scores (risks ranged from 80% to 89%); and G6: subgroup 6 of total absolute risk scores (risks ranged from 90% to 100%). The red dashed line is the population average. S1: stratum 1 of non-modifiable risk scores (risks ranged from 0 to 19%); S2: stratum 2 of non-modifiable risk scores (risks ranged from 20% to 39%); S3: stratum 3 of non-modifiable risk scores (risks ranged from 40% to 59%); S4: stratum 4 of non-modifiable risk scores (risks ranged from 60% to 79%); S5: stratum 5 of non-modifiable risk scores (risks ranged from 80% to 89%); and S6: stratum 6 of non-modifiable risk scores (risks ranged from 90% to 100%). The red dashed line is the population average.Sensitivity analyses were performed to investigate the robustness of the results, with b̑ calculated from previously published large cohorts or meta-analyses. First, the collected ORs for all risk factors were used to calculate absolute risks based on previously published cohort studies [Supplementary Table 3, https://links.lww.com/CM9/B760]. Second, the collected OR for smoking, the most significant modifiable lifestyle-related risk factor, was obtained from published meta-analyses [Supplementary Table 4, https://links.lww.com/CM9/B760]. Third, the bootstrap resampling method was performed to calculate the averaged ORs and b̑ for over 100 replicates. By applying the ORs collected from published studies and bootstrapping, the cumulative risks of esophageal cancer of the absolute risk model were similar to those of our main analysis [Supplementary Table 5 and Supplementary Figures 3–5, https://links.lww.com/CM9/B760]. Primary prevention by maintaining a good lifestyle can help provide basic information on decreasing the burden of esophageal cancer as well as practical guidelines for relevant cancer management. Our study can help develop cancer prevention and control strategies. Furthermore, assessing cancer risk and identifying the effects of modifiable and non-modifiable risk factors can reduce the utilization of medical resources in the hierarchical risk management strategy. In the present study, the results were derived from two population-based screening programs. Different from previous hospital-based studies, the population-based screening programs consecutively enrolled participants who were representative and had limited selection bias, which is the strength and social significance of our study. Our study results indicate potential targets for cancer management policies, demonstrating that policy-makers should consider encouraging residents to adopt a healthy lifestyle by modifying smoking and dietary habits to the lowest risk level to better allocate medical resources. By using modifiable risk factors in absolute risk modeling, we observed a marked risk reduction in esophageal cancer development. Our study showed that modifiable risk factors are responsible for more than 50% of esophageal cancer cases in the general population. Individuals who are at a higher risk from non-modifiable factors can gain more benefits through lifestyle modifications than those who are at a lower risk. This study provides insights into risk-based primary cancer prevention. However, the age range of 40–69 years, the geographic location, and the limited data on lifestyle information in this study are not representative of all populations in China and therefore limited the external validity. Funding This study was supported by grants from the National Natural Science Foundation of China (No. 72104150), MOE Project of Key Research Institute of Humanities and Social Sciences (No. 22JJD910001), CAMS Innovation Fund for Medical Sciences (No. 2021-I2M-1-010), the Natural Science Foundation of Beijing (No. 7204249), and Platform of Public Health & Disease Control and Prevention, Major Innovation & Planning Interdisciplinary Platform for the "Double-First Class" Initiative, Renmin University of China. Conflicts of interest None.