Introduction

Cancer and diabetes are two of the most significant health challenges globally, with both conditions exhibiting rapid increases in incidence, affecting global morbidity and mortality [1, 2]. The relationship between the two is complex, with evidence suggesting a bidirectional association. People with cancer are at an increased risk of diabetes [3,4,5], primarily due to the metabolic dysregulation driven by cancer and the consequences of cancer treatments [6,7,8]. Conversely, insulin resistance and elevated insulin levels promote tumor growth and impede programmed cell death [9, 10]. Importantly, this association is not just a concurrent health issue: diabetes in cancer survivors is linked with increased mortality rates compared to survivors without diabetes [11,12,13]. Moreover, for cancer patients, diabetes is associated with increased medical complications [14], a greater incidence of hospitalization [15], and lower health-related quality of life [16]. Thus, it is imperative to identify and manage modifiable risk factors for diabetes to mitigate the risk in this population.

Physical activity is beneficial to protect against the onset of diabetes. Observational studies [17,18,19,20] and randomized controlled trials [21,22,23,24] consistently demonstrate an inverse association between physical activity and the incidence of diabetes in the general population, underscoring the significance of physical activity as a key modifiable risk factor for diabetes. Nonetheless, no current research has examined the impact of physical activity on diabetes risk after a cancer diagnosis. While considerable evidence supports the benefits of physical activity for survival outcomes [25,26,27,28,29], its specific role in diabetes risk reduction among cancer survivors remains underexplored. Moreover, adherence to recommended physical activity level is notably low among cancer patients, with a substantial drop during and after treatment [30], despite guidelines promoting exercise for improved health outcomes [31, 32].

To bridge this knowledge gap, our nationwide study utilizes data from the Korean National Health Insurance Service (NHIS) to explore the association between changes in physical activity and the subsequent risk of diabetes among cancer survivors from various primary sites. We focus on the relationships between physical activity alteration before and after a cancer diagnosis with diabetes risk in this population.

Materials and methods

Database source

The NHIS operates as the sole insurance provider in Korea, delivering medical coverage to roughly 97% of the Korean population. It also oversees the provision of medical aid to those in the lowest income bracket.

The NHIS provides general health screening to all individuals aged 40 and above and employees of any age, who are eligible to participate in the national general health screening program at least once every 2 years at medical institutions throughout Korea [33]. The program includes anthropometric measurements, social and medical history questionnaires, and laboratory tests. A standardized questionnaire collects medical history and lifestyle behaviors such as smoking, alcohol consumption, and physical activity. Notably, the medical treatment database, which catalogs medical bills submitted by healthcare providers for reimbursement, can be cross-referenced with the health examination database. Therefore, the NHIS curates a wide-ranging health information dataset that spans the entire Korean population and that frequently has been utilized in epidemiological studies in Korea [34,35,36].

Study population

We identified a total of 351,767 individuals who were newly diagnosed with cancer between January 1, 2010, and December 31, 2016. All of these individuals participated in general health screening examinations within a 2-year period before and after their cancer diagnosis. We excluded 7794 individuals with missing or erroneous values in these examinations. We further excluded subjects aged < 20 (n = 3) and those with prior history of type 1 diabetes (n = 15,430), fasting plasma glucose level ≥ 126 mg/dL in general health screenings, or any history of type 2 diabetes (n = 57,620). After excluding incident diabetes within 1 year after cancer diagnosis (n = 6670), a total of 264,250 cancer survivors were identified and included in our analyses (Fig. 1).

Fig. 1
figure 1

Flowchart of the study population

Full size image

Cancer adjudication

A cancer diagnosis was established if a patient’s record contained both an International Classification of Diseases, Tenth Revision (ICD-10) code starting with “C” and a specific insurance claim code for cancer (V193). According to the policies of the NHIS, cancer patients are responsible for only 5% of their total medical expenses for cancer-related treatments, utilizing a unique co-payment reduction code (V193), which mandates a medical certificate from a physician. Therefore, the reliability of cancer diagnoses in this study is high, with a 97.9% sensitivity and 91.5% positive predictive value [37]. This method has been employed in prior studies [38, 39].

Ascertainment of physical activity changes

Information regarding physical activity was gathered through general health screenings before and after a cancer diagnosis using the modified International Physical Activity Questionnaire (IPAQ) [40]. Participants self-reported how many days during the preceding week they participated in light, moderate, or vigorous physical activity. The questionnaire provided an example of moderate physical activity, such as carrying light items, cycling at a steady place, or playing doubles tennis, and examples of vigorous activities that included heavy lifting, digging, aerobic exercises, or rapid cycling.

For this study, participants were categorized as either being adherent to physical activity, defined as engaging in a minimum of 30 min of moderate-intensity activity at least 5 days a week or at least 20 min of high-intensity activity at least 3 days a week, or non-adherent to physical activity [31]. Employing guideline adherence as the basis for classification provides a more precise evaluation of the influence of recommended physical activity levels on diabetes risk among cancer survivors [19], compared to quantifying physical activity in metabolic equivalent of task (MET) hours, as MET-based analysis is not feasible in our study setting with survey questionnaire. Four groups were constructed based on changes in physical activity status with respect to cancer diagnosis: remained inactive, became inactive, became active, and remained active.

Study outcome: diabetes

The primary outcome of this study was the incidence of newly diagnosed diabetes. Diabetes was defined by ICD-10 codes ranging from E11.x to E14.x, accompanied by the use of antidiabetic medications or a fasting glucose level of 126 mg/dL or higher. The cohort was followed from 1 year after the date of the post-cancer diagnosis general health screening examination to the date of incident diabetes, censored date, death, or the end of the study period (December 31, 2019), whichever came first. This approach was selected to exclude cases of pre-existing diabetes or temporary diabetes induced by cancer treatments (e.g., steroid use) and to allow a sufficient observation period post-treatment for the potential development of diabetes [41].

Covariates

Covariates were assessed at the post-diagnosis health screening examination. Age and income were recorded. Anthropometric measures were collected from general screening examinations. Obesity was defined following the Asian-Pacific criteria, with a body mass index (BMI) ≥ 25 kg/m2 considered obese [42]. BMI was calculated as weight in kilograms divided by the height in meters squared (kg/m2). Participants’ comorbidities were identified based on laboratory measures, claims, and prescription information prior to the index date as follows: hypertension (ICD-10 codes (I10.x-I13.x and I15.x), use of antihypertensive medication, or blood pressure ≥ 140/90 mmHg), dyslipidemia (ICD-10 code E78.x with lipid-lowering medication or total cholesterol ≥ 240 mg/dL), and chronic kidney disease (CKD; glomerular filtration rate < 60 mL/min/1.73 m2 as estimated by the Modification of Diet in Renal Disease equation). Information on smoking (current/no) and alcohol consumption (yes/no) was obtained from the general health screening after cancer diagnosis.

Statistical analyses

General characteristics are presented as means and standard deviations for continuous variables and as counts and percentages for categorical variables. To examine the significance of differences in proportions or means across four groups, chi-square tests were employed for categorical variables and one-way analysis of variance tests for continuous variables. The Fine-Gray proportional sub-distribution hazards model was used to calculate sub-distribution hazard ratios (sHRs) and 95% confidence intervals (CIs) for diabetes risk with death as a competing risk [43]. The proportional hazards assumption was assessed using Schoenfeld’s residuals, and no specific departure was observed. The reference group was “remained inactive,” and sHRs and 95% CIs were calculated for each group relative to the reference group. sHRs were obtained through a multi-step adjustment process. In the first model (Model 1), HRs were unadjusted. We identified potential confounders in the multivariable-adjusted models a priori based on a literature review [44]. Model 2 incorporated age, sex, income, smoking, alcohol consumption, obesity, hypertension, dyslipidemia, and CKD. In the final step (Model 3), we further adjusted for primary site of cancer. Subgroup analysis by primary site of cancer was performed using Model 2. Stratified analyses were conducted based on age, sex, and obesity-related cancer to identify interactions between changes in physical activity and diabetes risk. The definition of “obesity-related cancer” was followed to the International Agency for Research on Cancer (IARC) working group (Supplemental Table 1) [45]. Regarding breast cancer, we defined postmenopausal breast cancer as occurring at age 50 or older, considering the average age of menopause in Korea [46]. This definition was used because our current cohort data did not include information on menopausal status. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). P values provided are two-sided, and the level of significance was set to 0.05.

Ethics statement

This study was approved by the Institutional Review Board of Soongsil University (No. SSU-202303-HR-465–1). Anonymized and de-identified information was used for analyses, and informed consent was not required. The database is open to all researchers whose study protocols are approved by the official review committee.

Results

The cohort comprised 264,250 cancer survivors, with a mean age of 56.7 (12.5) years and 44.2% males. Among these survivors, 62.6% consistently remained inactive, while 9.8% consistently remained active. While 16.4% became active post-diagnosis, 11.2% became inactive post-diagnosis (Table 1). The “became inactive” group was the oldest, and the “became active” group was the youngest. Variations in waist circumference and the prevalence of obesity, current smoking, alcohol consumption, hypertension, dyslipidemia, and CKD among four groups were reported (all P < 0.001). However, these variations were considered clinically minimal.

Table 1 Baseline characteristics of the study population according to physical activity change
Full size table

Among primary sites of cancer, distinct patterns were observed in physical activity change patterns. Most cervical (70.5%), corpus uteri (64.9%), and skin cancer (68.6%) patients remained inactive. Notable shifts from inactive to active were observed in breast (23.4%), ovarian (20.3%), and Hodgkin’s lymphoma (21.3%) cases. A decrease in activity level was significant in prostate cancer (14.3%), whereas sustained physical activity was most common in thyroid (8.5%), testicular (9.8%), and corpus uteri cancer (8.2%) survivors.

Associations of physical activity change with diabetes risk after cancer diagnosis

During a mean follow-up period of 4.0 (2.0) years with 1,065,802 person-years, we observed 12,196 new cases of diabetes among cancer survivors (Table 2). In the sociodemographic-, traditional diabetes risk factor-, and primary site of cancer-adjusted model (Model 3), survivors with persistent physical activity had a 10% decreased risk of diabetes (sHR 0.90, 95% CI 0.85–0.96). Cancer survivors who became active or inactive after cancer diagnosis showed a slightly decreased risk of diabetes (sHR 0.98, 95% CI 0.93–1.03; sHR 0.97, 95% CI 0.92–1.03, respectively). Kaplan–Meier curves showing the estimated incidence probability of diabetes over time are presented in Fig. 2.

Table 2 Association of physical activity change with diabetes risk after cancer diagnosis
Full size table
Fig. 2
figure 2

Estimated incidence probability of diabetes after cancer diagnosis. Kaplan–Meier curves displaying the estimated incidence probability of diabetes by changes in physical activity

Full size image

Subgroup analyses by primary site of cancer

We examined associations between physical activity changes post-diagnosis and risk of diabetes among various cancer types (Table 3 and Fig. 3). For stomach cancer survivors, initiating physical activity post-diagnosis was associated with a 17% reduced risk of diabetes (sHR 0.83, 95% CI 0.71–0.96), while other activity patterns were only marginally associated with decreased diabetes risk. Similarly, among lymphoma survivors, post-diagnosis activity initiation was correlated with a 46% decrease in diabetes risk (sHR 0.54, 95% CI 0.32–0.91). In breast cancer survivors, a marginal 14% reduction in diabetes risk was observed with post-diagnosis physical activity (sHR 0.86, 95% CI 0.72–1.02), whereas lung cancer survivors showed a marginal 24% decrease in diabetes risk with sustained physical activity (sHR 0.76, 95% CI 0.57–1.01), without notable associations in other patterns. A similar trend was observed in survivors of liver and thyroid cancer.

Table 3 Subgroup analyses according to primary cancer site
Full size table
Fig. 3
figure 3

Sub-distribution hazard ratios (sHRs) and confidence intervals (CIs) of diabetes in various cancer sites

Full size image

Conversely, for pancreatic cancer survivors, changes in physical activity level post-diagnosis did not correlate with diabetes risk. For survivors of multiple myeloma, an increased risk of diabetes was noted across all three physical activity change patterns. However, the small number of events for survivors of these cancer types precluded any meaningful interpretation.

Stratified analyses according to age, sex, and obesity-related cancer

Stratified analyses showed no significant interactions of age, sex, and obesity-related cancer between changes in physical activity and diabetes risk among cancer survivors (Table 4).

Table 4 Stratified analyses based on age, sex, and obesity-related cancer
Full size table

Discussion

To the best of our knowledge, this is the first large-scale cohort study to investigate physical activity changes and risk of diabetes after cancer diagnosis. In our nationwide cohort of 264,250 survivors of cancer across all primary sites, regular physical activity maintained from pre- to post-diagnosis was associated with an overall decreased risk of diabetes. Physical activity either only before or only after cancer diagnosis showed slightly decreased risks of diabetes. The subgroup analyses demonstrated varied associations across cancer types.

By measuring physical activity repeatedly, we observed that sustaining regular physical activity from pre-diagnosis was associated with a 10% reduced risk of diabetes after cancer diagnosis. While previous research has predominantly assessed effects at a single time point, our findings reinforce the role of sustained physical activity on metabolic health, extending its known benefits to reducing the risk of diabetes after a cancer diagnosis. During adjuvant therapy, cancer patients often encounter unintentional weight gain, skeletal muscle loss, and increased insulin resistance [47, 48], which contribute to a higher risk of diabetes. Furthermore, corticosteroid administration during cancer management can cause hyperglycemia and subsequent onset of diabetes [49]. The risk is further exacerbated by the sedentary lifestyles of cancer patients, mostly related to the deconditioning effects of cancer treatment [50,51,52]. Physical activity plays a crucial role in this context, helping to mitigate these adverse effects by enhancing insulin sensitivity [53, 54], assisting in weight management [55], and promoting lean muscle mass [56], key factors affecting glycemic control.

There was only a slight and not significant risk reduction of diabetes in cancer survivors who became inactive after cancer diagnosis. Compared to survivors who maintained active lifestyles after cancer diagnosis, these inactive individuals appeared to benefit insufficiently from regular physical activity to prevent the development of diabetes. Although no strict formula can predict the precise amount or duration of physical activity necessary to prevent diabetes, long-term consistency is essential. Studies such as the Diabetes Prevention Program (DPP) demonstrated that lifestyle intervention can significantly reduce the risk of type 2 diabetes by 58% over a 3-year period [57], and follow-up studies such as the Diabetes Prevention Program Outcomes Study (DPPOS) have shown that these benefits were sustained over a 10-year period and beyond [58]. Another study, the Finnish Diabetes Prevention Study (DPS), followed participants for a median of 9 years and found that lifestyle intervention reduced the risk of type 2 diabetes by 33% [59]. These findings highlight the critical role of ongoing physical activity in diabetes prevention, a lesson of particular importance for cancer survivors who may deal with metabolic disturbances and deconditioning due to rigorous cancer treatments [50,51,52]. Therefore, the findings of our study emphasize the importance for cancer survivors to persist with a sufficient level of physical activity they had established prior to their cancer diagnosis as a strategic measure to reduce the heightened risk of diabetes following cancer treatment.

In our study, starting regular physical activity after cancer diagnosis was not associated with a significant reduction in diabetes risk. This subgroup (consisting 16.4% of our cohort) was characterized by the youngest average age and had the lowest prevalence of obesity, hypertension, and dyslipidemia following diagnosis—factors typically associated with lower diabetes risk. In addition, this group had the lowest rates of current smoking and alcohol consumption compared to the other groups in our study. The lack of observed benefit in terms of diabetes risk may be due to the relatively short duration of follow-up or possibly unmeasured confounding variables such as the specifics of exercise regimens (type, intensity, frequency, and timing), steroid use, and the use of immune checkpoint inhibitors. The influence of diet in conjunction with physical activity also warrants consideration, given its significant impact on metabolic health. To affirm the well-established association of physical activity with diabetes risk reduction through improved glycemic control, enhanced insulin sensitivity, and weight management among cancer survivors, further long-term observational and intervention studies are necessary.

The results of subgroup analysis indicate that the role of physical activity after a cancer diagnosis may differ according to the type of cancer. It is particularly noteworthy that stomach cancer survivors who began exercise after their diagnosis experienced a 17% decrease in the risk of diabetes. For lymphoma survivors, the decrease was even more significant, with a 46% reduction in risk. On the other hand, breast, lung, liver, and thyroid cancer survivors who either maintained or initiated physical activity post-diagnosis exhibited only marginal risk reductions, suggesting that the impact of physical activity on metabolic pathways can vary with the type of cancer. These differences could be attributable to the distinct treatment regimens for each primary site and variations in survivorship durations.

Limitations of our study include an observational study design that prevented causal inference and the measurement of physical activity by self-report questionnaire. The reliance on self-reported physical activity data can introduce recall bias, which could underestimate or overestimate the true association. In addition, the general health screening setting of our cohort could introduce selection bias, as it may not include individuals with severe health conditions. Moreover, the physical activity assessment was limited to two time points. Future studies might benefit from more frequent measurements or the use of pedometers for more accurate tracking. Last, information on cancer stage and treatment was not included in our cohort data.

Conclusions

Our findings suggest that sustaining regular physical activity from pre-diagnosis is associated with a lower risk of diabetes after a cancer diagnosis, independent of established diabetes risk factors. While associations between being physically active either only before or only after a cancer diagnosis and a lower risk of diabetes are suggestive, they are not statistically significant. Future research is warranted to establish clinical practice guidelines.