This retrospective study used secondary data from KNHANES to compare the health-related behaviors of middle-aged cancer survivors and non-cancer controls. This study was approved by the Institutional Review Board (IRB) of the Hwasun Chonnam National University Hospital (IRB No. CNUHH-2022-108). The dataset was officially approved as national statistics (approval No. 117002) and collected according to the National Health Promotion, Statistics, and Personal Information Protection Acts. KNHANES VI (2013–2015) and VII (2016–2018) data were analyzed. Out of 47,217 survey participants, only 17,332 individuals aged 40–65 years were included in this study. Participants who answered “don’t know” or refused to answer the questions related to cancer diagnosis (n=1,917) and health-related behaviors (n=2,232) were excluded from the study, resulting in a final sample size of 13,183 individuals. Among these individuals, only 750 had a history of cancer, with 12,433 individuals having no cancer history. To maintain a homogenous sample, individuals diagnosed within the past 2 years were excluded due to ongoing acute treatment and post-treatment effects, such as emotional stress, anxiety, depression, fatigue, and physical impairments [11]. Consequently, the final sample included 591 middle-aged cancer survivors diagnosed within the last 2 years (Fig. 1).

Study variables included sociodemographic characteristics, such as age, sex, marital status, education level, employment status, household income, and residential area, body mass index (BMI), comorbidities, activity limitations, stress level, and self-reported health status. According to the criteria established by the Korean Society for the Study of Obesity, the participants were categorized as underweight (<18.5 kg/m²), normal weight (18.5–24.9 kg/m²), or obese (≥25.0 kg/m²) based on their BMI [12]. The participants were further categorized as married or non-married (single, divorced, or separated) based on their marital status and as <high school, high school graduate, or ≥college graduate based on their education level. Additionally, the participants were classified as employed or unemployed based on their employment status and categorized into four groups (low, middle-low, middle-high, and high income) based on their household income. Based on their residential area, they were categorized as urban or rural. Comorbidities included physician-diagnosed conditions, such as hypertension, dyslipidemia, cardiovascular diseases, arthritis, chronic respiratory diseases, diabetes, thyroid diseases, depression, chronic kidney disease, and liver diseases (hepatitis B, hepatitis C, and cirrhosis). Activity limitations were assessed with the question, “Do you have any limitations in your daily life and social activities due to current health problems or physical/mental disabilities?” The responses were categorized as “Yes” or “No.” Stress level was assessed with the question, “How much stress do you usually experience in daily life?” The responses were categorized as none, mild, moderate, or severe. Self-reported health status was assessed with the question, “How would you rate your overall health?” The responses were categorized as good, fair, or poor.

Health-related behaviors were assessed based on the key lifestyle factors associated with cancer prevention and survival. Cancer screening adherence was defined as having undergone screening within the past 2 years, including both self-paid screening at comprehensive health centers and that provided by the National Health Insurance Corporation. Participant responses for screening adherence were categorized as “Yes” or “No.” Alcohol consumption was assessed based on the “monthly drinking rate” indicated in the KNHANES data, with participant responses classified as “Yes” if they consumed alcohol at least once per month or “No” if they were lifetime abstainers or consumed alcohol less than once per month. Based on their smoking status, the participants were classified as “current smokers” or “non-smokers/former smokers who quit smoking.” Physical activity was assessed according to the World Health Organization guidelines, with moderate-intensity activity defined as at least 150 minutes activity per week, vigorous-intensity activity as at least 75 minutes activity per week, or equivalent combination of both moderate and vigorous activities (1 minute of vigorous activity=2 minutes of moderate activity) [13]. Sedentary time was evaluated based on the self-reported daily sitting time, with prolonged sedentary time (>7 hr/day) classified as a risk factor for increased mortality [14]. Patient responses to the dietary control question, “Do you control your diet for special health reasons?” were categorized as “Yes” or “No.”

As the KNHANES data were collected using a complex sampling design rather than complete enumeration, survey weights, stratification variables, and cluster variables were incorporated into the statistical analyses. For the sociodemographic characteristics, unweighted numbers and weighted percentages were used as categorical variables, whereas means and standard errors were used as continuous variables.

Propensity score matching (PSM) was performed using a non-random package in R, version 4.1.0 to minimize the selection bias [15]. Each middle-aged cancer survivor was matched with five non-cancer controls based on variables potentially influencing the health-related behaviors, including age, sex, BMI, marital status, education level, employment status, household income, residential area, comorbidities, activity limitations, stress level, and self-reported health status. Univariate analyses were conducted using chi-square tests, whereas multivariate analyses were performed using binary logistic regression. Multivariate logistic regression models were developed using two approaches to evaluate the health-related behaviors before and after PSM. Model 1 was adjusted for PSM and significant variables (P<0.1) identified via the univariate analyses of health-related behaviors, whereas Model 2 was adjusted for PSM and all covariates. Two-tailed P-value <0.05 was considered statistically significant. All statistical analyses, except PSM, were performed using SPSS ver. 28.0 (IBM Corp.).

Patient characteristics before and after PSM are presented in Table 1. Among the 13,024 participants, mean age of the 591 cancer survivors, including 136 males (26.3%) and 455 females (73.7%), was 53.8 years. Before matching, mean age of 12,433 non-cancer participants, including 5,071 males (47.8%) and 7,362 females (52.2%), was 51.2 years. After 1:5 PSM, the final sample comprised 591 cancer survivors and 2,955 non-cancer controls. Prior to matching, significant differences in baseline characteristics, including age, sex, BMI, education level, employment status, household income, comorbidities, activity limitations, and self-reported health status, were observed between both groups (P<0.05). However, after matching, no significant difference in any characteristic was observed between both groups (P>0.05), confirming successful matching.

Among the 591 middle-aged cancer survivors, thyroid cancer (n=152; 25.7%) was the most common, followed by breast (n=114; 19.3%), other (n=95; 16.1%), cervical (n=78; 13.2%), stomach (n=76, 12.9%), colon (n=39; 6.6%), liver (n=10; 1.7%), and lung (n=10; 1.7%) cancers. Notably, 17 participants (2.9%) were diagnosed with multiple primary cancers. The time elapsed since diagnosis was ≥2 years but <5 years in 201 participants (34.0%) and ≥5 years in 390 participants (66.0%) (Table 2).

Health-related behaviors of both groups before and after PSM are presented in Table 3. After matching, significant differences in alcohol consumption, physical activity, and diet were observed between both groups. In total, 217 middle-aged cancer survivors (39.0%) consumed alcohol compared to the 1,325 controls (48.5%) (P<0.001). Moreover, 288 cancer survivors (50.7%) engaged in moderate or vigorous physical activity, and 1,276 controls (44.0%) adhered to the recommended physical activity guidelines (P=0.011). Dietary management was significantly different between both groups, with only 211 cancer survivors (35.8%) showing diet adherence compared to the 797 controls (26.5%) (P<0.001). However, no significant differences in cancer screening adherence, smoking status, and sedentary time were noted between both groups.

Next, multivariate logistic regression analyses were conducted to examine the differences in health-related behaviors between middle-aged cancer survivors and non-cancer controls (Table 4). In Model 1, which was adjusted for PSM and significant covariates (P<0.1) determined via univariate analyses, middle-aged cancer survivors were significantly less likely to consume alcohol (odds ratio [OR]: 0.64; 95% confidence interval [CI]: 0.51–0.81) and smoke (OR: 0.64; 95% CI: 0.42–0.97) but more likely to engage in physical activity (OR: 1.33; 95% CI: 1.08–1.63) and practice dietary management (OR: 1.47; 95% CI: 1.17–1.83) than the non-cancer controls. No significant differences in cancer screening adherence and sedentary time were observed between both groups.

Model 2, which was adjusted for PSM and all covariates, showed the same results, with middle-aged cancer survivors exhibiting lower odds of alcohol consumption (OR: 0.64; 95% CI: 0.52–0.81) and smoking (OR: 0.65; 95% CI: 0.43–0.99) and higher odds of engaging in physical activity (OR: 1.33; 95% CI: 1.08–1.64) and practicing dietary management (OR: 1.47; 95% CI: 1.17–1.84) than the non-cancer controls. However, no significant differences in cancer screening adherence and sedentary time were noted between both groups. All logistic regression models incorporated sampling weights to account for the complex survey design of KNHANES.