Our study was the first to identify the trajectories of PA frequency and its relationship with all cancer risk and several specific cancers. Our findings showed that more than two-thirds of middle-aged Korean adults remained at a low frequency of PA, and only 5% had a high frequency of PA during approximately 7 years. Additionally, this study revealed that a small proportion of people changed their frequency of PA from low to high and from high to low, and each trajectory accounted for approximately only 5% in both sexes. Existing evidence suggests that PA has a protective effect on cancer prevention, and PA in almost studies was usually measured at a single time point (i.e., baseline) [19,20,21,22,23]. Hence, we hypothesized that PA trajectories during the 7 years could have modified the association between PA and cancer incidence. The present study unveiled a novel finding that, compared to persistent low frequency, maintaining a high frequency of PA over a period of approximately 7 years was significantly associated with a lower risk of all cancer incidence among women.

The link between PA and cancer risk, especially that of specific cancers, has been established in observational studies; however, to our knowledge, no study has assessed the effect of PA trajectories on the risk for all cancers. Limited research has shown an inverse impact of non-trajectory-based PA on the development of all cancers [24, 25]. In particular, a systematic review and meta-analysis of 47 studies involving 5,797,768 participants and 55,162 cases showed a decreased risk for digestive-system cancers in the high PA level group (Relative risk (RR) = 0.82, 95%CI = 0.79–0.85) compared to the low PA level group, and significant results were observed in both sexes [24]. A study pooling 1.44 million adults from 12 prospective US and European cohorts also indicated that a high level of baseline leisure-time PA was associated with lower risks for 13 cancers [25]. Additionally, the significant effect of high-to-low trajectory PA on cancer risk was not observed in our study, thereby emphasizing the importance of maintaining a high frequency of PA during a long period for the reduction of the risk for all cancers, rather than performing a high frequency of PA temporarily. However, the proportion of people with the persistent high trajectory of PA accounted for approximately 5%; therefore, promoting daily PA is crucial for preventing cancer development, especially in women. Although our study could not demonstrate the protective impact of the low-to-high trajectory of PA on cancer risk, we believed that if a high level of PA was maintained for a long duration, rather than persistently low PA, the risk for cancer development could be reduced. Therefore, further studies are required to assess PA trajectories over longer period to confirm its association with cancer risk. Our study also showed a favorable effect of PA on cancer prevention in women only after adjusting for potential covariates, such as smoking and alcohol consumption. In fact, men are more likely to use tobacco and drink alcohol than women, as also observed in our study, and both behaviors are the most important risk factors attributable to cancer development. Additionally, the higher cancer susceptibility among men due to exposure to carcinogens from work, unwillingness to seek healthcare or sex-related biologic factors could be latent factors [26], which could not be controlled in our study. Therefore, sex differences in lifestyle behaviors could explain the negligible effect of PA on all cancer risk in men observed in our study.

World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) determined the relationship of total PA with lung cancer risk to be limited-suggestive evidence [2]. A recent meta-analysis of 20 cohort studies has demonstrated that increased PA was associated with a lower risk for lung cancer in both sexes and in smokers, but not in non-smokers [27], and this finding was consistent with other previous studies [28, 29]. In our study, we observed the pronounced association between lung cancer and a moderate level of PA in both smokers and non-smokers, but in men only. The heterogeneity among studies related to the link between lung cancer and PA level, including measurement of PA, study design, and study participants’ characteristics, is challenging, making the results of the numerous studies not comparable. Our study measured the frequency of general PA, while others focused on either recreational PA, non-recreational PA, or all domains of PA. The underlying mechanisms of PA can prevent lung cancer development through several pathways, including reducing insulin resistance and inflammation, decreasing oxidative stress and enhancing DNA repair mechanisms, increasing enzymatic systems and cofactors such as glutathione that detoxify chemical carcinogens, and enhancing the innate and acquired immune response [2]. Furthermore, the protective impact of PA on lung cancer prevention was seen in former smokers and non-smokers in our study. This could probably be attributed to the endogenous antioxidant defenses in the association between PA and lung cancer, as seen in the molecular epidemiology study within EPIC project [20]. Moreover, the relatively low number of lung cancer cases among women could reduce the statistical robustness that could interfere with the statistical significance observed in our study. Given that smoking is the most important factor attributable to lung cancer, prevention of smoking initiation and promotion of smoking cessation is the most effective method of primary prevention against lung cancer. Additionally, our study suggested that actively engaging in PA enormously contributes to lung cancer prevention in men, regardless of their smoking status.

The association between PA and thyroid cancer is inconclusive. The very first case-control study in the US established the hypothesis that regular recreational exercise reduced the risk for thyroid cancer (Odds ratio (OR) = 0.76, 95%CI = 0.59–0.98) and was also supported by other case-control studies in European countries [23]. Another case-control study in South Italy also suggested that walking every day for at least 60 min diminished the risk for thyroid cancer development (OR = 0.357, 95%CI = 0.157–0.673). However, evidence from large cohorts does not support this hypothesis. A study in the US involving 484,326 men and women revealed that the risk for thyroid cancer and its subtypes was unassociated with vigorous exercise [30]. Another large study pooling data from five prospective studies which measured different aspects of PA (i.e., frequency of vigorous activities, metabolic equivalent task, number of hours spent performing vigorous activities, or strenuous exercise) showed that all patterns of PA were insignificantly associated with thyroid cancer development [31]. A null association between leisure-time PA and thyroid cancer was also observed in a large cohort study involving 1.4 million adults from the US and European countries, and the relationship between PA and cancer risk was not modified by BMI or smoking status [25]. In contrast, our study emphasized that the risk for thyroid cancer was reduced among men who engaged in low-to-high, high-to-low, and persistently high trajectories of PA frequency in comparison with men who engaged in a persistently low trajectory of PA frequency. The inconsistent methods of measuring PA and its domains, volumes, and time periods across epidemiological studies make the results difficult to compare among studies. Sufficient evidence of a consistent method to determine the true association between PA and thyroid cancer is currently unavailable. Furthermore, evidence showed that the risk for thyroid cancer was strongly associated with obesity in men [32]; additionally, our study highlighted the effect of a higher frequency of PA in reducing thyroid cancer risk in overweight and obese individuals. Based on our study results and the fact that thyroid cancer is one of the five most common cancers in Korea [33], we highlighted the need for high frequency PA over a long period, especially among overweight and obese people.

The favorable impact of PA on breast cancer prevention was observed in postmenopausal women, and limited-suggestive evidence for this has been shown among premenopausal women by the WCRF/AICR [2]. A meta-analysis of 38 cohort studies investigated the inverse link of PA to breast cancer risk in both premenopausal and postmenopausal women, and this association was consistent among all domains of PA [34]. In line with previous studies, the reduction in breast cancer risk was observed in women who had a persistently high frequency of PA in our study, compared with those who had a low frequency of PA. This finding is similar to that reported in another large cohort study [25]. However, we could not examine the beneficial effects of PA against breast cancer in premenopausal and postmenopausal women separately due to the lack of such information; therefore, future research is required to investigate the confounding effects of women-health variables, including menopause, parity, and hormone replacement therapy, on the relationship between trajectories of PA and breast cancer risk. Regular PA via a diverse array of mechanisms, such as reduction in circulating estrogens levels, insulin resistance, and inflammation, was reported to have a protective impact against breast cancer [2]. We could not observe the significant link between PA trajectory and breast cancer risk in the subgroup analyses by BMI. This could be due to the relatively small sub-population.

Convincing evidence demonstrated that the highest level of total PA reduced the risk for colon cancer by 20% versus the lowest level (RR = 0.80, 95%CI = 0.72–0.88) [2]; in contrast, PA had a negligible effect on the prevention of colorectal cancer in our study. A significant inverse association between PA and colorectal cancer in men did not remain after adjusting for potential confounders, such as alcohol consumption, smoking, BMI, and CCI. As for women, a null association between PA and colorectal cancer was seen in the present study, consistent with 13 other studies with RRs ranging from 0.69 to 1.15, whereas another study showed an increase of PA between baseline and follow-up showed the beneficial effect of PA on the prevention of colon cancer only, not of rectal cancer and combined colorectal cancer [35]. In the subgroup analysis, we observed a decrease in the risk for colorectal cancer in male smokers who maintained a moderate level of PA, similar to another large cohort study that measured the effect of PA on risk for colon and rectal cancer independently [25]. Obesity is another established risk factor for colorectal cancer, especially in men [36]. In our study, moderate PA was insufficient to lower the risk for colorectal cancer in overweight/obese men, even though significant results were seen in underweight and normal men. Based on this finding, we placed a strong emphasis on weight management in conjunction with PA to maximize cancer risk prevention.

Findings regarding the impact of PA on endometrial cancer risk in women have been largely equivocal as some studies showed an inverse association [37, 38], while others reported no link [39], similar to the findings of our study. However, it is notable that a higher risk for corpus uteri cancer was associated with the persistently high frequency PA group in underweight and normal BMI persons. To the best of our knowledge, this finding has not been previously reported in the literature. In contrast, other studies indicated that high lifetime PA was linked to a higher risk for endometrial cancer in overweight and obese women [38]. We hypothesized that a lack of women-health covariates in our study could modify the effect of PA on cancer risk. In addition, given the limited sample size in sub-categories, larger studies that are adjusted for sufficient covariates will be needed to clarify the role of BMI in the relationship between PA and corpus uteri cancer before firm conclusions can be drawn.

Additionally, while our study observed a null association between PA trajectory and risk for liver cancer, WCRF/AICR showed evidence that the association between liver cancer and PA was “limited-suggestive” [25], and a meta-analysis demonstrated an inverse relationship between daily total PA and liver cancer risk [40]. The discrepancies in the measurement of various PA domains and time periods could confound the relationship of PA with the risk for these cancers. Therefore, a sufficient number of studies with homogeneous methods of PA assessment are needed to confirm the true association of PA with the risk for liver cancer.

Our study had several limitations. First, since the NHIS cohort did not use the global PA questionnaire developed by the WHO [41], we could not collect detailed information regarding certain domains of PA and did not measure the metabolic equivalents of the tasks. For example, people who did PA regularly between 60 and 70% of their target heart rate may report that they did not sweat during, which may underestimate the effect of PA on cancer prevention. Second, Due to only including participants who received health examinations, this cohort study was unable to represent the entire Korean population. Furthermore, the fact that there were twice as many men as women can be attributed to the fact that most of the men were employed and the head of family, who were eligible for the general health screening program. Meanwhile, only individuals aged 40 years and older who were the dependents of the employed person and head of household were eligible for the program, and the majority of them were women. It also contributed to the fact that women were older than men in our study [10]. Third, the cancer incidence in this study was defined based on ICD-10 codes in primary diagnosis and a special code for verifying cancer in the claims data; however, a secondary diagnosis of cancer was not considered. This could lead to an underestimation of cancer incidence. Forth, despite the enormous number of participants, the small number of incident cases of rare cancer in the trajectory strata could reduce the statistical power. Finally, the study lacked an additional set of covariates significant in developing specific cancer sites (e.g., menopause status in relation to breast cancer), which could potentially confound the association of PA with cancer risk. Therefore, cancer-specific covariates should be considered as confounding factors in further analyses.