PDF
Citation
Print
Introduction
Previous studies have shown that the risk of diabetes mellitus (DM) is higher in cancer survivors compared with populations without cancer [1-3]. In two recent meta-analyses, DM risk was increased by approximately 1.4-fold among cancer survivors [2,3], which is an important issue as DM can negatively affect disease outcomes and the general health of cancer survivors. Hyperglycemia and high insulin level in DM patients might promote cancer cell growth and result in increased cancer mortality and recurrence rate [4]. DM and its complications, such as renal insufficiency, might also limit the choice and intensity of cancer therapy and impede treatment [4]. In addition, DM is an established risk factor of cardiovascular disease and is reported to cause a 2- to 4-fold increase in risk [5]. Cardiovascular disease is a common cause of death among aged cancer survivors and overtakes primary cancer mortality over time from cancer diagnosis [6,7]. Therefore, the identification of high-risk groups for incident DM after a cancer diagnosis is needed for further preventive schemes.
Obesity, which has become a global epidemic, contributes to the development of DM [5,8]. Obesity is associated with numerous types of cancers, including post-menopausal breast cancer, colorectal cancer, and endometrial cancer [9], and the prevalence of obesity among cancer survivors is increasing [10]. Nevertheless, evidence is lacking on the relationship between obesity status and incidence of DM in cancer survivors. Also, body weight might change after a cancer diagnosis [11], but research is insufficient regarding weight change in various cancer types. Here, our population-based study investigated the associations between weight change after a cancer diagnosis with incident DM, both in overall cancers and specific cancer types.
Materials and Methods
1. Data source and study population
This population-based cohort study used data from the Korean National Health Insurance Service (NHIS). The NHIS is the single insurer in Korea and provides almost the entire population with universal health coverage. The NHIS database includes information on demographics and medical records of the Korean population including diagnosis codes, drug prescriptions, and medical institution usage. The NHIS also provides its beneficiaries 40 years of age or older or who are employees with a national health screening examination, biennially or annually. The national health screening data contain information from health questionnaires about past medical history and lifestyle, laboratory tests, and anthropometric measurements [12].
From this database, 1,413,422 individuals who were newly diagnosed with cancer between 2010 and 2016 were identified. From them, 351,767 individuals who received a national health screening within 2 years before and after a cancer diagnosis were included. Individuals who had a missing value in health screening data (n=7,794), who were aged under 20 years (n=3), who were previously diagnosed with type 1 DM (n=15,430), and/or who had prior diagnosis of type 2 DM or who showed a fasting glucose level ≥ 126 mg/dL in the national health screening (n=57,620) were excluded sequentially. Individuals who were diagnosed with incident DM within a 1-year lag period from cancer diagnosis were washed out (n=6,670). The final study population included 264,250 cancer survivors (Fig. 1).
2. Definition of cancer diagnosis
A diagnosis of cancer was defined by the claim records with relevant International Classification of Diseases-10th revision (ICD-10) diagnosis codes (S1 Table) in conjunction with a co-payment reduction program code for cancer (V193). The special co-payment reduction program of the NHIS is intended to support its subscribers diagnosed with critical illness by subsidizing medical expenses. Patients who have a medical certificate of cancer diagnosis issued by a physician and who registered for this program pay only 5% of their medical expenses related to cancer. Therefore, almost all Korean cancer patients are enrolled in this program, which assures the reliability of definition of cancer diagnosis used in our study [13].
3. Exposure: weight change
Weight change was defined based on anthropometric measurements from the national health screening examination, which is performed according to a standard measurement protocol [12], before and after a cancer diagnosis. Body mass index (BMI) at each health screening was calculated by dividing body weight (kg) by the square of height (m2). Then, weight status was categorized into obese (BMI ≥ 25 kg/m2) or normal weight (BMI < 25 kg/m2) according to the Asia-Pacific criteria of the World Health Organization [14], which are commonly accepted in Korea [15]. Based on weight status before and after a cancer diagnosis, weight change of study subjects was categorized into four groups: (1) sustained normal weight, (2) obese to normal weight, (3) normal weight to obese, or (4) sustained obese.
4. Covariates
Information about demographics (sex, age, and income) at baseline (a health screening examination after a cancer diagnosis) were extracted from the NHIS database. Income level was dichotomized at the lowest quartile.
Lifestyle data (smoking, alcohol consumption, and regular physical activity) were collected from health questionnaires at the baseline national health screening. Subjects who reported alcohol intake > 0 g/day were defined as drinkers. Performing moderate physical activity ≥ 5 days/wk or vigorous physical activity ≥ 3 days/wk was defined as regular physical activity.
Comorbidities were defined based on anthropometric measurements and laboratory tests at the baseline national health screening and prescription records. Hypertension was defined as prescription records of antihypertensive medications with relevant ICD-10 codes (I10-I13 and I15), systolic blood pressure ≥ 140 mmHg, or diastolic blood pressure ≥ 90 mmHg. Dyslipidemia was defined as prescription records of lipid-lowering medications with relevant ICD-10 code (E78) or serum total cholesterol level ≥ 240 mg/dL. Chronic kidney disease was defined as an estimated glomerular filtration rate (eGFR) based on the Modification of Diet in Renal Disease equation < 60 mL/min/1.73 m2.
5. Study outcome
The main outcome of our study was incident DM after a cancer diagnosis. DM was identified by healthcare utilization under relevant ICD-10 codes (E11-E14) with prescription records of antihyperglycemic medications. Study subjects were followed from the index date (1 year after the health screening examination post-cancer diagnosis) to the date of incident DM, death, or December 31, 2019, whichever came first.
6. Statistical analysis
Baseline characteristics of the study population according to weight change were determined by descriptive analyses. One-way analysis of variance tests (ANOVA) were used for comparisons of continuous variables, and chi-square tests were used for comparisons of categorical variables. Cox proportional hazard analyses were performed to examine adjusted hazard ratios (aHR) and 95% confidence intervals (95% CI) of DM incidence according to weight change, using a sustained normal weight group as the reference. Proportional hazard assumption was evaluated before the analyses using Schoenfeld residuals. Model 1 was unadjusted, model 2 was adjusted for potential confounders selected based on previous studies (age, sex, income, smoking, alcohol consumption, regular physical activity, hypertension, dyslipidemia, and chronic kidney disease) [1,8], and model 3 was additionally adjusted for cancer type [3]. Subgroup analyses were also performed on differences in main outcome according to cancer type. In addition, we conducted competing risk analysis regarding death as a competing risk using the Fine-Gray proportional subdistribution hazards model.
All statistical analyses were performed using SAS ver. 9.4 (SAS Institute Inc.). Two-tailed p < 0.05 were considered statistically significant.
Results
1. Weight change and baseline characteristics of the study population
Among the final study population of 264,250 cancer survivors, 163,395 (61.8%) sustained normal weight, 19,558 (7.4%) changed from obese to normal weight, 14,625 (5.5%) changed from normal weight to obese, and 66,672 (25.2%) sustained obese after a cancer diagnosis. Subjects showed different weight change patterns according to cancer type, but the proportions of the sustained normal weight group were highest among the weight change groups in all cancers (S2 Table).
Baseline characteristics of the study population at health screening after a cancer diagnosis are presented in Table 1. Subjects who sustained normal weight were more likely to be female and younger than other weight change groups. Patients in the obese to normal weight group were less likely to have low income, smoke, or drink than other groups, while they performed more regular physical activity than other groups. Patients in the sustained obese group had more comorbidities (hypertension, dyslipidemia, or chronic kidney disease) and higher BMI, waist circumference, glucose level, and systolic and diastolic blood pressure and lower eGFR than other groups, whereas the contrary result was shown in the sustained normal weight group (all p < 0.001).
2. Weight change after a cancer diagnosis and risk of DM in overall cancers
Over a mean 4.0 years (standard deviation, ±2.0) of follow-up, a total of 12,196 cancer survivors (4.6%) was newly diagnosed with DM. Risk of DM was highest in the sustained obese group, followed by the normal weight to obese group, the obese to normal weight group, and then the sustained normal weight group in all three models. In fully adjusted Model 3, the aHR and 95% CI were 2.17 (2.08-2.26) for the sustained obese group, 1.66 (1.54-1.79) for the normal weight to obese group, and 1.29 (1.21-1.39) for the obese to normal weight group (Table 2). Competing risk analyses showed similar associations between weight change and incident DM to the main analyses (S3 Table).
3. Weight change after a cancer diagnosis and risk of DM according to cancer type
In subgroup analyses according to cancer type, most of the cancers showed the highest aHRs in the sustained obese group (Table 3). Bladder, breast, cervix, colorectum, corpus uteri, lung, oral cavity, stomach, thyroid, and other cancers showed similar association patterns to overall cancers. However, small sample size and event numbers made it difficult to confirm significant differences according to weight change group in several cancers.
Discussion
In this large population-based study of cancer survivors, weight status after a cancer diagnosis was associated with risk of incident DM. DM risk was highest in patients in the sustained obese group, followed by the normal weight to obese group, the obese to normal weight group, and then the sustained normal weight group. Similar patterns were observed between weight change and DM risk in most cancer types, although it was not statistically significant for some cancers.
The literature has reported that weight change may occur after diagnosis of several types of cancers [11,16,17]. For instance, breast cancer survivors are known to experience weight gain frequently, whereas weight loss is frequent in upper gastrointestinal tract cancer survivors [11,17]. However, evidence is insufficient regarding weight change patterns in survivors of various types of cancers. Also, most previous studies were conducted in Western populations and studies regarding weight change among Asian cancer survivors, particularly Korean survivors, are limited in number. A few Korean studies, which mostly included a small number of breast cancer patients, showed different weight change patterns from Western studies and suggested that ethnic differences might exist regarding weight change after cancer diagnosis [18,19]. Therefore, our large-scale investigation on Korean cancer survivors is expected to provide valuable data.
In our study, 87% of the study population maintained their weight status, 7.4% changed from obese to normal weight, and 5.5% changed from normal weight to obese after a cancer diagnosis. The proportion of obese participants slightly decreased from 32.6% (pre-cancer diagnosis) to 30.8% (post-cancer diagnosis), while the general population at this age (mean 56.7 years) usually shows an increase in obesity prevalence over time [20]. Several previous investigations on the relationship between cancer diagnosis and weight change have suggested that obese cancer survivors are more likely to lose weight than normal weight survivors or obese cancer-free controls, possibly due to the effort to reduce weight or adopt healthy behaviors, at least partly [18,21]. Our finding that patients in the obese to normal weight group were more likely to participate in regular physical activity and less likely to smoke or drink than other weight change groups after a cancer diagnosis seems to support this hypothesis. Obesity is suggested to increase the risk of recurrence or mortality in some cancers, which might affect the intention to lose weight among cancer survivors [9,18]. Nevertheless, a majority (77.3%) of obese cancer survivors sustained their weight status in our study, and 8.2% of normal weight survivors became obese after a cancer diagnosis.
Among overall cancer survivors, obesity at pre- and/or post-diagnosis was associated with increased risk of incident DM, presenting a 2.2-fold increased risk after full adjustment in the sustained obese group compared with the sustained normal weight group. Obesity defined by BMI is a well-known modifiable risk factor of DM, and the relative risk of type 2 DM is reported as increased by 2.6-fold in the obese population of Korea, which is similar to our finding [8,22]. Insulin resistance is postulated to play a key role in the development of DM among obese individuals [23]. Increased secretion of nonesterified fatty acids and other substances like adiponectin or cytokines from adipose tissue can contribute to insulin resistance. Impaired pancreatic β-cell function in obesity can also result in dysregulation of glucose levels and consequent development of DM [23].
Cancer survivors who changed from obese to normal weight had far lower risk of DM (aHR, 1.29; 95% CI, 1.21 to 1.39) than survivors who sustained obesity (aHR, 2.17; 95% CI, 2.08 to 2.26). Evidence is lacking regarding the relationship between weight loss and DM incidence in obese cancer survivors. A small study of overweight breast cancer survivors reported a decrease in insulin resistance among participants who lost ≥ 5% of their body weight through behavioral intervention [24]. Previous research in the general population has consistently suggested the protective effect of intentional weight loss. In prospective studies of the overweight or obese population at risk, interventions for lifestyle modification and subsequent weight reduction resulted in significantly reduced risk of DM [25,26]. A Diabetes Prevention Program study of 3,200 individuals reported that participants who were assigned to lifestyle intervention, which was targeted to achieve at least 7% weight loss, lost an average 5.6 kg of body weight and showed a 58% reduction of DM incidence compared to the placebo group over a 3-year follow-up period [26]. Our finding implies that increased DM risk from obesity can be substantially decreased, and a similar protective effect against DM can be expected from weight reduction in obese cancer survivors as in the general population. To date, most trials on weight loss interventions in obese cancer patients have focused on breast cancer and showed that the interventions might be effective and safe [27,28]. Future research is warranted regarding the effects of weight loss intervention on weight reduction and DM prevention in various cancers.
One notable finding of this study is that the risk of DM was higher in the normal weight to obese group (aHR, 1.66; 95% CI, 1.54 to 1.79) than in the sustained normal weight group (reference). This is important because survivors of certain cancer types, such as breast cancer and endometrial cancer who underwent hysterectomy, might commonly experience weight gain after a cancer diagnosis [11,29]. Especially, breast cancer survivors are well-known to frequently gain weight, possibly due to chemotherapy and changes in lifestyle [11]. As our study has shown a considerable increase in DM risk among patients in the normal weight to obese group, close monitoring of weight change and early effort to prevent obesity during and after a cancer diagnosis might be necessary in these high-risk groups.
Subgroup analyses according to cancer type showed similar patterns to the main analysis for many cancers, and most of the other cancers also showed the highest aHRs in the sustained obese groups. However, it was hard to attain statistical significance, likely due to the small sample size and event numbers in many cancer types. Nevertheless, this finding implies that obesity can increase the risk of DM in survivors of various cancers. To date, most of the weight management strategies for cancer survivors have focused on weight increase or prevention of unintentional weight loss, except for those with cancers well-known to associate with obesity [30]. Our finding suggests that more active obesity management is needed in overall cancer survivors to prevent DM.
This study has important clinical implications. As mounting evidence has shown that obesity in cancer survivors is related to poor outcomes and general health, the importance of weight management has been suggested by experts [31,32]. The American Cancer Society has stated that there is no contraindication to modest weight reduction in obese cancer survivors even when they are undergoing cancer treatment [31]. However, multiple barriers seem to exist that impede weight loss in cancer survivors [33,34]. Most healthcare providers including oncologists tend to focus only on cancer treatment and are indifferent to weight management, resulting in a lack of advice regarding weight and inadequate shared care. Some healthcare providers consider obesity only as a risk factor of poor cancer outcome and are unconcerned with comorbidity risk, which might weaken the motivation to help patients achieve healthy weight after cancer treatment. Referral options to nutrition or exercise experts may also be insufficient. Further, intervention programs for weight management in cancer survivors are rare despite high patient concerns, and some new drugs for obesity are not yet established as safe for cancer survivors. Economic problems and geographical distance might also limit the access to weight loss interventions [33,34]. Our findings that obesity at any time point (before or after cancer diagnosis) is associated with DM risk and that weight changes after cancer diagnosis might alter the risk support the significance of continuous weight management among cancer survivors, including weight reduction in obese survivors and obesity prevention in normal weight survivors. As the prevalence of DM is drastically increasing worldwide [5], systematic efforts to enhance the resources and strategies for weight management in cancer survivors are needed. Healthcare providers should be aware of the importance of maintaining or achieving normal weight in cancer survivors and consider weight loss intervention for obese survivors regardless of the time point.
Our investigation has several limitations. First, selection bias might exist as we included cancer survivors who received national health screening. However, our study population might have been a more clinically relevant population to obesity management who had better general health condition than individuals who did not participate in health screening. Second, as we used an established database, we could not obtain some clinical information that might affect the results, including cancer stage and prognosis, treatment modality (surgery, chemotherapy, or hormone therapy), or disability. However, our national database provided us with a large representative sample population. The intent to lose weight was also not investigated, while unintentional weight loss can occur among cancer patients, especially in upper gastrointestinal tract cancer patients due to poor gastrointestinal function or decrease in dietary intake [16,17]. Nevertheless, our findings show that weight loss, whether intentional or unintentional, may help prevention of DM. As our study population was limited to those who participated in health screening twice, the participants were expected to be mostly in good general health, limiting the possibility of unintentional weight loss among patients of cancers other than gastrointestinal tract cancers. In addition, we did not distinguish underweight from normal weight patients. As the number of underweight patients is expected to be small, it is unlikely to significantly affect the study results. Finally, we could not obtain more detailed weight change history between baseline and post-diagnosis health screening.
Nevertheless, our study has specific strengths. To our knowledge, this is the first study to examine associations between changes in body weight and DM risk of overall cancer survivors. Previous studies were mostly on survivors of specific cancer types, and data on Asian ethnicity, especially Korean, is lacking. Only a few studies with a small number of specific cancer types have suggested the possibility of change in insulin resistance or DM risk according to weight change [24]. Our nationwide research in Korea provided evidence in this population regarding relationship between incident DM risks with weight gain or weight loss after diagnosis of overall cancers. As we used a large population-based cohort, we could also investigate the differences according to various cancer type.
In conclusion, obesity at any time point was associated with increased risk of DM in cancer survivors, presenting the highest risk in cancer survivors who sustained obesity. Cancer survivors who changed from obese to normal weight had a lower risk of DM than survivors who sustained obesity. Survivors who changed from normal weight to obese showed increased risk relative to those who sustained normal weight. Weight management should be emphasized in cancer survivors, and specific strategies targeted for this population are needed.
XML Download