Introduction
Type 2 diabetes mellitus (DM) is a metabolic disorder characterised by chronic hyperglycaemia due to progressive loss of pancreatic β-cell function in insulin secretion on the background of insulin resistance and metabolic dysfunction.1 Previous meta-analyses have shown that patients with diabetes are at higher risk of developing cancer at several sites, including the liver,2 pancreas,3 colon and rectum,4 bladder5 and possibly kidney.6 With the exception of bladder cancer, the remaining four cancers have also been linked to obesity.7 Patients with obesity are 3–7 times likely to develop diabetes,8 and both conditions are characterised by metabolic dysfunction. However, beyond obesity and hyperglycaemia, metabolic dysfunction is recognised to be present when other metabolic abnormalities (such as lipid abnormalities and hypertension) exist. While high triglycerides or low high-density lipoprotein cholesterol (HDL-C) are widely accepted as lipid abnormalities,9 consensus on lower limit for triglycerides or upper limit for HDL-C has yet to be reached. Nevertheless, metabolic dysfunction is linked to the five common cancers of diabetes regardless of the presence or absence of obesity even though there is a lack of evidence for defining metabolic abnormalities at a level contributing to an elevated risk of cancer. In addition, while bladder cancer has not been shown to be obesity related7 but is common among patients with diabetes5 and metabolic dysfunction,10 gastric cancer is one of the obesity-related cancers11 and its linkage to diabetes12 and metabolic dysfunction13 has not been demonstrated.
While epidemiological research has identified that the same types of obesity-related cancers were found among patients with diabetes and those who suffered from metabolic dysfunction, there is no systematic examination of which metabolic dysfunction components, precipitating behaviours, and comorbid chronic illnesses, are linked to which types of obesity-related cancers among patients with diabetes. Nevertheless, prior research suggests that obesity-related cancers at different organ sites are characterised by different profiles of metabolic dysfunction and clinical characteristics. For example, while obesity has been linked to a number of cancer sites such as the colon and rectum, liver, pancreas, kidney and stomach, strong evidence only exists for the links between body mass index (BMI), but not other adiposity indicators and different cancer sites.7 Moreover, while diabetes has been shown to be associated with various cancer sites such as the colon and rectum, liver, pancreas and bladder, glycaemic levels such as glycated haemoglobin (HbA1c) or fasting blood glucose are only independently associated with the risk of liver14 and pancreatic cancers.15 16 Furthermore, while low triglycerides and high HDL-C are known to protect against atherosclerotic cardiovascular diseases, it remains unknown whether lipid abnormalities in triglycerides or HDL-C are associated with the risk of cancer at different organ sites. In addition, while hypertension is known to be linked to an increased risk of kidney cancer,17 limited evidence provides support to the links between hypertension and other cancer sites. Also, while liver cirrhosis is known to increase the risk of liver cancer, it remains largely unknown how it affects the risk of extrahepatic malignancies.18 Similarly, the relationships between chronic kidney disease (CKD) and cancer sites other than kidney are inadequately understood. Furthermore, some observational studies showed that metformin,19 statins,20 aspirin or non-steroidal anti-inflammatory drugs (NSAIDs)21 are potentially protective against cancer, however, findings remain inconsistent19–22 and their roles in site-specific cancer risk remain less clear.
To fill the research gap on the different profiles of metabolic dysfunction among patients with diabetes who develop different site-specific obesity-related cancers, the current study aims to examine the cancer outcomes of patients with diabetes who received an assessment of metabolic dysfunction, precipitating behaviours and comorbid chronic illnesses in public general outpatient clinics (GOPCs) of Hong Kong.
Materials and methods
Study design and setting
This is a retrospective cohort study based on territory-wide electronic health records of Hong Kong’s public healthcare system. The hospital authority (HA) is a statutory body providing 90% of inpatient care and 30% of outpatient services (including both specialist and primary care levels) to the general public. Currently, the HA manages 43 public hospitals, 49 specialist outpatient clinics and 74 GOPCs over the territory. The HA maintains a centralised repository of patient demographics, prescription records, clinical diagnoses, outpatient attendances, inpatient admissions and laboratory results. Disease diagnoses were coded according to the International Classification of Disease 9th or 10th revision (ICD-9 or ICD-10) or the International Classification of Primary Care second edition. Data were accessed via Hospital Authority Data Collaboration Lab.
Patients
Patients who received DM diagnosis and referred for DM complication screening (DMCS) at one of the GOPCs were initially included. Those with (1) non-type 2 DM; (2) missing date of DM diagnosis; (3) DM diagnosis below the age of 18 years; (4) history of malignancy or (5) a first assessment prior to 2010 (as over 95% of patients received a first assessment between 2010 and 2019) were excluded. In addition, those who (1) received primary cancer diagnosis not belonging to any of the six target sites; (2) received primary cancer diagnoses at more than one target sites or (3) were followed up for less than 6 months were also excluded. Patients were assessed for their clinical profile during an initial DMCS assessment and followed up until a cancer diagnosis, death or December 2019, whichever earlier. Figure 1 shows the flow chart of patient selection.
Flow chart of patient selection. DM, diabetes mellitus; DMCS, diabetes mellitus complication screening.
Measures
Independent variables included metabolic dysfunction indicators in four categories: (1) obesity (BMI and waist-to-hip ratio); (2) insulin resistance or impaired glucose tolerance (HbA1c and fasting glucose); (3) serum lipid profile (low-density lipoprotein cholesterol (LDL-C), HDL-C and triglycerides) and (4) hypertension. Other variables included demographics (sex and age), duration of diabetes, lifestyle behaviours (alcohol consumption and smoking), disease history (liver cirrhosis, ischaemic heart disease, cerebrovascular disease, heart failure, chronic obstructive pulmonary disease, pneumonia, CKD, family history of diabetes), medication use (antidiabetic drugs, statins, aspirin, NSAIDs, anticoagulants, antiplatelets and antihypertensive drugs) and laboratory measures (serum creatinine). CKD is defined as estimated glomerular filtration rate G3a–G5 or urine albumin-to-creatinine ratio A2–A3. For liver cancer, chronic hepatitis B (CHB) and chronic hepatitis C (CHC) were also included.
The outcomes of interest were an inpatient diagnosis of cancer at the colon and rectum (ICD-9: 153–154; ICD-10: C18–21), liver (ICD-9: 155; ICD-10: C22), pancreas (ICD-9: 157; ICD-10: C25), bladder (ICD-9: 188; ICD-10: C67), kidney (ICD-9: 189; ICD-10: C64–66, C68) and stomach (ICD-9: 151; ICD-10: C16). Epidemiological evidence suggests that liver,2 pancreatic,3 colorectal4 and bladder5 cancers are more common among the diabetes population than among the general population. Some evidence also supports the association between kidney cancer and diabetes.6 On the other hand, liver, pancreatic, colorectal, kidney and gastric cancers are linked to obesity.7 Patients were separated into seven mutually exclusive groups, which are six site-specific cancer groups and one reference group that did not receive any cancer diagnosis during follow-up.
Data analysis
Patient characteristics by site-specific cancer outcomes were presented in mean with SD or median with IQR for continuous variables, and in count with proportion for categorical variables. Cox proportional hazards regression was applied to examine the relationships between a set of covariates and each site-specific cancer outcome among patients who developed a site-specific cancer and those who remained free of cancer during follow-up. In each model, a full set of covariates (metabolic dysfunction indicators, demographics, duration of diabetes, lifestyle behaviours, disease history, medication use and biochemical measurements) were included to mutually control the effects of each other. Estimates of the covariates in the models were reported in adjusted HR with 95% CI. Model performance was measured using Harrell’s concordance (C-) index as metric. As a post hoc analysis, mean value of individual lipid component against the number of years before death (as a proxy for underlying disease severity) was plotted by cancer group. All analyses were performed by using R V.3.5.2 (R Foundation for Statistical Computing, Vienna, Austria). Missing values were handled using replacement by attribute mean of the same class. No adjustment was made for multiple comparisons. Statistical significance was set at p<0.05, two sided.
Patient and public involvement
None.
Discussion
The current study is among the first to examine the differential contribution of a comprehensive profile of metabolic dysfunction and other clinical factors to site-specific cancer risk among patients with diabetes. The findings of the study revealed that a profile of metabolic dysfunction indicators (obesity, insulin resistance and serum lipids), behavioural (smoking), disease history and medication use may differentiate the subsequent site-specific risk of obesity-related cancers, including cancers of the colon and rectum, liver, pancreas, kidney and stomach, as well as bladder cancer, among metabolically compromised individuals.
The present study demonstrated that insulin resistance or impaired glucose tolerance, as indicated by elevated HbA1c and fasting glucose levels, is associated with an increased risk of liver and pancreatic cancers, but not other four studied cancer sites. While previous meta-analyses supported the positive links between fasting glucose and the risk of liver14 and pancreatic15 cancers across the range of pre-diabetes and diabetes, a UK Biobank study found that HbA1c is only independently linked to the risk of pancreatic cancer but not other 15 cancers among the general population, suggesting the limited role of hyperglycaemia in promoting carcinogenesis of different organ sites.16 Findings of the current study are generally consistent with the literature supporting the stronger relationships between glycaemic indicators and pancreatic cancer as well as a potential association between glycaemic level and liver cancer among the diabetes population. This could be partially explained by the unique roles of the pancreas and liver in insulin production and responses to insulin signalling, respectively, in glucose homeostasis.23 It has also been proposed that since cancer cells are highly effective in glucose uptake, hyperinsulinaemia, rather than hyperglycaemia itself, could be more critical to inducing carcinogenesis.23
Moreover, the present study demonstrated that lower lipid levels are generally associated with higher cancer risk at several sites, except the positive link between HDL-C and liver cancer. While existing literature on the associations between individual lipid components and site-specific cancers is scarce and findings remain inconsistent,24 25 some proposed that the inverse links between various lipid components and cancer risk prior to cancer diagnosis could be due to high metabolic demands of subclinical tumour cells during growth and proliferation.26 Another possible mechanism is upregulation of PDL-1 in adipose tissue may promote tumour growth via suppressing immune function, but attenuate obesity-associated chronic inflammation and insulin resistance, hence potentially reducing circulating lipid levels.27 28 Post hoc analyses of the study also suggested that under normal healthy state (no cancer), low HDL-C and high triglycerides are associated with poor health outcomes (death) as expected. However, lipid levels exhibit different patterns among patients who eventually develop cancer. The altered lipid patterns could be potentially due to underlying tumour metabolism or changing adipose tissue functioning, implying that lipid profiles alone may not be adequate indicators of cancer risk among metabolically compromised patients.
In addition, the distinctive positive association between HDL-C and liver cancer could be due to the unique role of liver in cholesterol homoeostasis and HDL functionality. While HDL is known for its antioxidative and anti-inflammatory properties to protect against cardiovascular diseases via reverse cholesterol transport, its association with cancer risk has been rarely explored. Some studies supported the inverse link between HDL-C and risk of cancer.24 25 One possible explanation for the observed positive link between HDL-C and liver cancer risk is as hepatic tumour cells are highly upregulated in scavenger receptor class B type I expression to recruit HDL-C from peripheral tissues, circulating HDL-C may increase to fuel the growth and proliferation of tumour cells in the liver.29 On the other hand, it is also possible that dysfunctional HDL may have lost its atheroprotective function during hepatic carcinogenesis. Specifically, HDL family represents a heterogeneous group of HDL particles.30 31 As HDL subfractions continuously undergo interconversion, impaired remodelling may result in compositions which favour inflammation, promoting atherosclerosis and carcinogenesis.32 It has been previously reported that patients with both high HDL-C and high C reactive protein levels are at higher risk of recurrent coronary events,32 suggesting the potential inflammatory state indicated by an uncommonly high level of HDL-C.
The findings of the study showed that among patients with diabetes, adiposity is positively associated with the risk of colorectal, liver and bladder cancers, while hypertension is not linked to any of the six studied cancer sites. While epidemiological evidence supports the associations between obesity with a number of cancers, including gastric, colorectal, liver, pancreatic and kidney cancers,7 the inconsistent relationships between obesity and cancer sites in the current study could be attributed to differences in confounding control or less variation in adiposity among the diabetes population. On the other hand, prior research has demonstrated the association between hypertension and kidney cancer but not other cancer sites.17 The lack of association between hypertension and kidney cancer in the present study could be due to the almost ubiquitous presence of hypertension (prevalence over 85%) among the study population.
As expected, the current study demonstrated that smoking is associated with a number of cancers, including colorectal, pancreatic, liver and bladder cancers. The International Agency for Research on Cancer has classified tobacco smoking as a group I carcinogenic agent to humans for all six studied cancer sites.33 However, no association between smoking with kidney or gastric cancer has been observed in this study. Tobacco smoke is rich in carcinogenic compounds, such as polycyclic aromatic hydrocarbons, heterocyclic compounds (such as furan), N-nitrosamines and aromatic amines.34 Some carcinogens have been identified to be carcinogenic to specific cancer sites. For example, animal studies have demonstrated that furan is a liver carcinogen.34 In both animal and human studies, a tobacco-specific compound, N-nitrosamines 4-(methylnitrosamino)−1-(3-pyridyl)−1-butanone, is carcinogenic to the liver and pancreas, while aromatic amines, 4-aminobiphenyl and 2-naphthylamine, are carcinogenic to the bladder.34 The major route of carcinogenic pathways of tobacco smoke is formation of covalent bonds between carcinogens and DNA generating DNA adducts, leading to somatic mutations and eventually cancer.34
Furthermore, the present study showed that chronic diseases such as liver diseases, heart failure and CKD could be associated with some site-specific cancers. The findings of the study on chronic diseases and cancer risk are largely consistent with existing literature. This study showed that liver cirrhosis is not only associated with an increased risk of liver cancer but also extrahepatic malignancies such as colorectal and pancreatic cancers. While epidemiological evidence on the associations between cirrhosis and extrahepatic malignancies remains scarce, a previous study conducted in Sweden found that cirrhosis is linked to a 5.1-fold and 3.6-fold risk of pancreatic and colorectal cancers, respectively, among the general population.18 The exact pathophysiological mechanism remains unknown. One possible mechanism is that systemic chronic low-grade inflammation in cirrhosis may induce extrahepatic malignancies.35 Emerging evidence suggests that upregulation of GATA3 may promote adipose tissue inflammation, insulin resistance, fatty liver disease associated with metabolic dysfunction and liver cirrhosis,36 intensifying systematic inflammation and potentially accelerating development of cancer beyond the liver. Furthermore, the observed association between heart failure and elevated risk of liver cancer could be due to reduced cardiac output from cardiac dysfunction since under normal state, the liver receives approximately 25% cardiac output.37 Moreover, the current study showed that CKD is associated with an increased risk of liver and bladder cancers. The association between CKD and liver cancer could be due to the potential presence of metabolic dysfunction-associated renal dysfunction and fatty liver disease,38 while the link between renal dysfunction and bladder cancer could be due to accumulation of uraemic toxins in the urinary system.38
Moreover, the current study demonstrated the associations between different medication use and site-specific cancers. Results of the study are in general consistent with existing literature. This study showed that metformin use was linked to a reduced risk of kidney cancer. Although metformin use has been shown to be linked to a decreased risk of cancer, with the strongest evidence on colorectal cancer,39 evidence for its potential chemopreventive effect against kidney cancer remains limited.40 The proposed mechanisms are adenosine monophosphate-activated protein kinase activation and inhibition of glucose/insulin production.41 It is possible that metformin use may lower the risk of kidney cancer through glycaemic control and prevention of renal complications and kidney diseases. On the other hand, the current study showed that sulfonylurea use is associated with an increased risk of colorectal, gastric and liver cancers, while insulin use is associated with an elevated risk of liver, gastric and kidney cancers, but a lower risk of colorectal cancer. Previous epidemiological studies mostly compared use of sulfonylurea and insulin with use of other antidiabetic drugs such as metformin.23 While some studies observed an increased risk of cancer among sulfonylurea or insulin users, it is unclear whether the excess risk is genuine or due to reduced risk of metformin use.23 One explanation of the observed elevated risk of sulfonylurea or insulin use is overstimulation of pancreatic β-cells or direct exogenous insulin resulting in excess circulating insulin. Excess insulin may promote cancer cell proliferation due to overexpression of insulin and insulin-like growth factor (IGF) receptors in cancer cells and increased levels of circulating bioactive IGF-I.23 It is unclear whether the observed protective effects of insulin are due to differences in the effects of short-acting, intermediate-acting, and long-acting insulin, or counteractive effects from other medication use. On the other hand, use of statins, aspirin and NSAIDs consistently demonstrated protective effects against cancer development for several cancer sites. Previous observational studies generally supported the cancer protective effects of these few drugs.19–21 Pathophysiologically, statins may exert its antitumour effects through anti-inflammatory, proapoptotic, antiproliferative and antifibrotic properties,42 while aspirin and non-aspirin NSAIDs may exercise their antitumour effects through anti-inflammatory, proapoptotic and antiangiogenic properties.43 However, despite some promising results in observational studies, randomised clinical trials failed to demonstrate evidence to support repurposing of these drugs.19–21 On the other hand, despite experimental studies showing the antitumour effects of anticoagulants,44 45 the observed increased risk of colorectal and gastric cancers for anticoagulants use could be due to potential gastrointestinal bleeding.46 In addition, the observed positive link between use of antihypertensive drugs and kidney cancer is consistent with the literature, which suggests the positive associations between each class of antihypertensive drugs and kidney cancer.47
There are several limitations in the current study. First, this study only captured the profiles of patients with diabetes at baseline. However, their clinical profiles may change over the course of diabetes. Second, heavy alcohol use, hormonal, occupational and dietary information were not available in this study. These factors may confound the observed relationships between exposure and cancer risk. Third, due to the nature of retrospective study, and the fact that liver and pancreatic cancers may induce diabetes, the possibility of reverse causality cannot be completely ruled out.14 15 However, to minimise potential bias, patients who had a prior cancer diagnosis or received cancer diagnosis within 6 months of the assessment were excluded from analyses. Lastly, generalisability of the findings could be limited to older Chinese patients with type 2 DM.
This post was originally published on https://bmjopen.bmj.com