Introduction

Numerous epidemiological studies have indicated strong positive associations between type 2 diabetes mellitus (T2DM) and the incidence of several cancer types [1, 2], and mortality [3, 4]. The strength of association depends on cancer site. The most considerable associations have been demonstrated for pancreatic cancers [5,6,7], and liver [8], but there are observations of increased risks of colorectal [9], breast [10], endometrial [11], and kidney cancers [12]. A reduced risk of prostate cancer in diabetic males have documented in many studies. This effect has noticed especially strong among diabetic men with high body mass index (BMI) [13, 14].

The connection between diabetes and cancer has been explored extensively, but the observed associations are not completely understood. It has been suggested, that possible underlying mechanisms could associate with insulin resistance and its accompanying hyperinsulinemia, hyperglycemia, and chronic inflammation [8, 15, 16]. Hormonal factors, low testosterone levels in particular, have been suggested to explain the reduced risk for prostate cancer among male diabetics [17].

In recent years, attention has also been paid on the potential role of several anti-diabetic drugs on cancer risk [18,19,20], but the results are inconsistent. Some studies but not all have suggested that insulin analogs may increase the cancer risk [20], and that the biguanide metformin apparently reduces the risk of many cancer types [21, 22], but these findings have not confirmed in all studies, and causality has not been proven.

The aim of this study was to describe the cancer risk pattern among the Finnish T2DM patients. Special attention was paid to the cancers with a priori suspected associations with diabetes.

Materials and methods

Persons diagnosed with diabetes in Finland during 1988–2007 were identified from the nationwide FinDM II database, which comprises persons with diabetes identified from: (1) the Register of Persons Eligible for Special Reimbursement of Medication for chronic conditions including diabetes (years 1964–2007); (2) the Prescription Register including all reimbursed medicines purchased from the Finnish pharmacies (years 1994–2007); (3) the National Hospital Discharge Registers including all inpatient care (years 1969–2007) and outpatient hospital visits (years 1998–2007); (4) the Causes-of-Death Register (1971–2007); and (5) the Medical Birth Register (1987–2007). The unique Finnish personal identity code assigned to all residents in Finland allowed the deterministic record linkage. More detailed description of the study population is available elsewhere [23].

Persons were considered to have diabetes if they had started antidiabetic medication or had been hospitalized with a diagnosis of diabetes. We excluded from our analyses women with gestational diabetes only and people, who were not permanent residents in the country. The high reimbursement level for antidiabetic medication has resulted in a high coverage of diabetic persons in the medication registers. The completeness of the Hospital Discharge Register is confirmed by the law for every hospital in the country to report on the inpatient episodes. A very good coverage of diabetic patients in the FinDM II database was demonstrated in a study, where the database was compared to a local Diabetes Register of the Helsinki metropolitan area; only some diabetic persons aged 65 years or more treated in outpatient primary care settings without antidiabetic medication were missing [24].

The precise etiological classification into type 1 (T1DM) and T2DM was not possible with the current register-based data. Therefore, we determined the diabetes type based on age at diagnosis of diabetes: persons diagnosed before the age of 30 years were considered to have T1DM and the rest T2DM.

The follow-up for cancer incidence through the files of the national Finnish Cancer Registry and for vital status and emigration through the Population Register was performed using the personal identity code as the key. The follow-up for cancer incidence started at the date of diagnosis of the diabetes or on 1st of January 1989, whichever was later, and ended at death or emigration, or on December 31, 2014, whichever occurred first. The follow-up of 204,724 male and 223,602 female T2DM patients produced altogether 4.48 million person-years (Table 1). The mean length of follow-up of a person was 10.5 years.

Table 1 Number of type 2 diabetic patients under follow-up (N) and number of person-years at follow-up up to 31 December 2014, by sex and age
Full size table

The numbers of observed cases and person-years at risk were counted by 5-year age groups, separately for calendar periods (1989–1994, 1995–1999, 2000–2004, 2005–2009 and 2010–2014). The expected numbers of cases for total cancer and specific cancer types were calculated by multiplying the number of person-years in each age group by the corresponding average cancer incidence in all of Finland during the period of observation. The specific cancer types selected a priori for the analysis included the cancer sites with known or suspected exceptional (increased or decreased) risk in T2DM patients in earlier studies, and other common cancer types to give the whole picture of the cancer situation among Finnish T2DM patients.

To calculate the standardized incidence ratio (SIR), the observed number of cases was divided by the expected number. The 95% confidence intervals (CI) for the SIR were based on the assumption that the number of observed cases followed a Poisson distribution.

Results

During the follow-up, 74,063 cases of cancer were diagnosed; the expected number was 63,870. Consequently, the total cancer incidence for adults with T2DM was 16% higher than expected (SIR = 1.16, CI = 1.15–1.16; Table 2).

Table 2 Observed (Obs) and expected (Exp) numbers of cancer cases and standardized incidence ratios (SIR) with 95% confidence intervals (95% CI) among Finnish type 2 diabetic patients in 1989–2014, by site according International Classification of Diseases (ICD-10)
Full size table

The incidence of 16 cancer types was statistically significantly elevated; cancer of lip, stomach, colon, rectum, liver (SIR = 2.44, CI = 2.35–2.53), gallbladder, pancreas (SIR = 1.75, CI = 1.70–1.79), unspecified intestinal tract, non-melanoma skin, kidney, bladder, brain and central nervous system, thyroid, non-Hodgkin lymphoma, leukaemia, and ill-defined or unknown (Tables 2, 3).

Table 3 Observed (Obs) and expected (Exp) numbers of cancer cases and standardized incidence ratios (SIR) with 95% confidence intervals (95% CI) among Finnish type 2 diabetic patients in 1989–2014, by sex and site according International Classification of Diseases (ICD-10)
Full size table

For women with T2DM, 34,199 cancer diagnoses were observed, which was 18% higher than expected. We found statistically significant increase for cancers of lip, stomach, colon, rectum, liver (SIR = 1.90, CI = 1.77–2.03), gallbladder, pancreas (SIR = 1.72, CI = 1.65–1.78), unspecified intestinal tract, non-melanoma skin, kidney, non-Hodgkin lymphoma, corpus uteri, and other female genital organs (Table 3).

Among men with T2DM, 39,864 cancer diagnoses were observed, which was 14% more than expected. Statistically significant increase in cancer incidence were seen for cancers of lip, oral cavity, stomach, colon, rectum, liver (SIR = 2.79, CI = 2.66–2.92), gallbladder, pancreas (SIR = 1.78, CI = 1.71–1.85), unspecified intestinal tract, non-melanoma skin, kidney, bladder, thyroid, and other male genital organs (C60, C63). The inverse effect was seen for prostate cancer only (SIR = 0.95, CI = 0.93–0.96); 10,692 men with T2DM developed prostate cancer, which was 5% less than expected (Table 3).

Discussion

The total cancer incidence among the 428,326 T2DM patients was 16% higher among people with T2DM compared with the expected average cancer incidence in the Finnish population. The increase in cancer incidence was seen both in women (18%) and men (14%).

We found an almost threefold increase in liver cancer incidence among men with T2DM and nearly twofold increase among women with T2DM compared with the general population. Similar findings have been observed in several studies earlier [1, 8, 25]. It has been speculated whether diabetes is a direct risk factor for liver cancer or whether there are other common risk factors involved, such as non-alcoholic fatty liver disease (NAFLD), steatosis or cirrhosis [26, 27].

For pancreatic cancer we observed a more than 70% increased risk in both men and women with T2DM. This result applies to what is already known about the relationship between T2DM and cancer of the pancreas [5,6,7]. Suggestions of the possible underlying mechanisms between diabetes and pancreatic cancer include elevated insulin concentrations acting as a tumor-growth promoting factor in pancreatic cells, hyperglycemia, insulin resistance and hyperinsulinemia [5, 16]. It also has been questioned whether the association is caused by reverse causality and diabetes is caused by liver or pancreatic cancer [4]. However, this does probably not explain increased risks in our data because of the temporal relations. Pancreatic cancer is generally known as quickly progressive and fatal, elevated incidence may be due to reverse causality only in the very short time frame (less than 6 months) following diabetes onset. Sometimes pancreatic cancer is misdiagnosed as diabetes, due to similar symptoms. Continuous increased risk of pancreatic cancer after longer periods of follow-up (up to 10 years after diabetes onset) is unlikely due to reverse causality [7, 15].

We found an increased risk of cancer of urinary organs, particularly kidney cancer, in persons with T2DM. This finding is similar to a meta-analysis of studies investigating the association between diabetes and kidney cancer incidence and mortality [12]. The association has been tried to be described by common risk factors, such as obesity and cigarette smoking, or medication. Influence of possible confounding or detection bias can not put aside, neither. Because diabetic patients are under increased medical surveillance, and the measurement of urinary specimens is recommended (mainly for detection of signs of diabetic nephropathy) it is possible that kidney cancer tend to be found at an earlier stage in diabetic patients than in non-diabetic people [28].

Although higher risk of breast cancer among type 2 diabetic patients has been indicated in many studies [10, 29], the connection was not seen in our data. Instead, we found an increase in cancers of female reproductive organs, such as cancer of corpus uteri, also known as endometrial cancer (50%), and cancers of vulva and vagina in diabetic women (24%). Endometrial cancer risk has reported to be higher among women with diabetes [11, 30], but this relationship may be confounded by body weight [31]. Nevertheless, a positive association between T2DM and endometrial cancer has still been seen after adjusting for possible confounding factors: physical inactivity and obesity [11].

In contrast to the increased risk for many types of cancer, a 14% decreased risk of prostate cancer has been seen in pooled analysis of 45 studies [13]. In our study the diabetic males possessed 5% reduction in prostate cancer risk. This inverse relationship is not fully understood, but has been considered to result from lower testosterone levels in diabetic patients, as well as many more hormonal and metabolic factors related to lifestyle and medication. It has been suggested that diabetic men are more likely to be tested for prostate-specific antigen (PSA) levels, and this lead to an increased probability of detecting prostate cancer, opposite to the observed inverse association [14].

Cancer registration system in Finland is virtually complete nationwide and the computerized record linkage procedure based on the national personal identifier is precise [32]. Therefore, technical incompleteness of case ascertainment does not cause any bias in our results. However, confounding by lifestyle factors, such as physical inactivity, dietary habits and cigarette smoking, is possible [33]. The overall cancer incidence among working-aged Finnish men increases and among women decreases towards the lower social class [34]. Low socioeconomic status (SES) has been associated with late state diagnoses and T2DM is more common in people with lower SES, as well [35].

Other strengths of the study are, besides technical accuracy, relatively long-term follow-up and size of the data, due to record linkage. Finnish registers provide access to valuable data without time-consuming and expensive data collecting. Observed and expected frequencies of cancers are come from the same database. Most studies report associations between diabetes and cancer mortality. Compared to mortality, incidence is, nevertheless, a preferable indicator for cancer risk due to variation in cancer survival rates due to earlier diagnosis and better treatment.

Many studies investigating the association between diabetes and cancer incidence have not made a proper distinction between the two main types of diabetes. The precise etiological classification between T1DM and T2DM was not possible with our current register-based data, which might have led to misclassification, where group of T2DM includes persons who actually had other type of diabetes, but such a probability is small compared with the large number of people with T2DM [36].

It is very challenging to study associations between these heterogenous and complex diseases. T2DM shares several common risk factors with many cancer types, such as obesity, poor diet, ageing and physical inactivity. Thus confounding is likely to occur and the possibility of reverse causality could not be ignored, either. Risk factors for T2DM are similar to cardiovascular diseases (CVD), as well. Thus, a long follow-up shows negative correlation with cancer mortality, because of diabetic patients’ high mortality on CVD [37]. Additionally, it has been speculated that if cancer is more common in diabetic patients, the relation between diabetes and cancer might be underestimated since T2DM, in fact, is underdiagnosed disease. Also, due to possible undiagnosed individuals with T2DM the cancer risk in the non-diabetic population might be higher than observed [8].

It is still unclear, whether T2DM and its metabolic derangements (hyperglycemia, insulin resistance and hyperinsulinaemia) increase the risk of cancer directly or is the association indirect. There is a wide array of possible mechanisms related with cancer incidence in patients with T2DM. Due to the multiple and complex mitogenic effect of insulin, hyperinsulinaemia is a potential factor favouring cancer progression [38]. Increasing insulin levels activate the related insulin-like growth factor-I (IGF-I) that probably promotes cancer cell growth and proliferation. Impaired glucose tolerance is related to increased cancer risk without diabetes, too. Poor metabolic control leads to increasing oxidative stress causing a chronic pro-inflammatory state, resulting vulnerable cells predisposing to malignant transformation. Biological actions, common with degenerative diseases, such as T2DM, might accelerate carcinogenic processes [16].

We need more in-depth, well-designed follow-up studies to advance our knowledge of the complex association between T2DM and cancer incidence, and the underlying direct and indirect biological mechanisms. It is essential to explore the potential role of glucose-lowering treatments. The task is challenging, because of the diversity of therapies over time. A series of confounders and biases should be observed properly to ensure validity. Because of the growing, worldwide burden of diabetes, it is extremely important to acquire understanding to these questions to prevent cancer in diabetic patients.