Cancer incidence in persons with type 1 diabetes: a five-country study of 9,000 cancers in type 1 diabetic individuals

Aims/hypothesis An excess cancer incidence of 20–25% has been identified among persons with diabetes, most of whom have type 2 diabetes. We aimed to describe the association between type 1 diabetes and cancer incidence. Methods Persons with type 1 diabetes were identified from five nationwide diabetes registers: Australia (2000–2008), Denmark (1995–2014), Finland (1972–2012), Scotland (1995–2012) and Sweden (1987–2012). Linkage to national cancer registries provided the numbers of incident cancers in people with type 1 diabetes and in the general population. We used Poisson models with adjustment for age and date of follow up to estimate hazard ratios for total and site-specific cancers. Results A total of 9,149 cancers occurred among persons with type 1 diabetes in 3.9 million person-years. The median age at cancer diagnosis was 51.1 years (interquartile range 43.5–59.5). The hazard ratios (HRs) (95% CIs) associated with type 1 diabetes for all cancers combined were 1.01 (0.98, 1.04) among men and 1.07 (1.04, 1.10) among women. HRs were increased for cancer of the stomach (men, HR 1.23 [1.04, 1.46]; women, HR 1.78 [1.49, 2.13]), liver (men, HR 2.00 [1.67, 2.40]; women, HR 1.55 [1.14, 2.10]), pancreas (men, HR 1.53 [1.30, 1.79]; women, HR 1.25 [1.02,1.53]), endometrium (HR 1.42 [1.27, 1.58]) and kidney (men, HR 1.30 [1.12, 1.49]; women, HR 1.47 [1.23, 1.77]). Reduced HRs were found for cancer of the prostate (HR 0.56 [0.51, 0.61]) and breast (HR 0.90 [0.85, 0.94]). HRs declined with increasing diabetes duration. Conclusion Type 1 diabetes was associated with differences in the risk of several common cancers; the strength of these associations varied with the duration of diabetes. Electronic supplementary material The online version of this article (doi:10.1007/s00125-016-3884-9) contains peer-reviewed but unedited supplementary material, which is available to authorised users.


Introduction
Details of each country's source of diabetes and cancer registrations follow below.

Nordic Countries
Population-level data on cancer incidence for Denmark, Sweden and Finland were extracted from NORDCAN [1], a multinational registry of cancer cases in each of the Nordic countries (Denmark, Sweden, Finland, Iceland, Norway and the Faroe Islands). This registry was formed in 2003 and is populated by data from each country's cancer registry.

Denmark
The Danish National Diabetes Register was established in 2006 [2] and was based on three existing health care registers, the Danish Civil Registration System, the National Patient Register and the National Health Service Register. As of January 1995, the Danish National Diabetes Register was considered complete with regards to incident cases. The Danish Cancer Registry [3] was established in 1943 and cancer reporting to this registry has been mandatory since 1987. A system of multiple notifications from several data sources including the Danish Pathology Register has ensured that complete coverage of cancer diagnoses has been attained. Data from these registers were linked using the Central Person Register number, a unique patient identifier assigned to every person in Denmark as of 1968 [4]. The period of follow-up was January 1995 until December 2012, a total of 18 years. In addition to patients diagnosed with type 1 diabetes after 1995, prevalent cases who were diagnosed before 1995 but were under 40 years of age in 1995, and therefore diagnosed below 40 years of age were also included in this study.
No ethical approval is needed for register based studies in Denmark. Approval for linkage was obtained from the Danish Data Protection Board.

Finland
Data on diabetes diagnoses were obtained from FinDM, a research database maintained by the National Institute for Health and Welfare (THL) [5]. This database contains all people identified to have a diabetes record since 1964 in a number of health care registries including hospital admissions and prescription databases. These nationwide health care registers are held by the National Institute for Health and Welfare (THL), the Social Insurance Institution of Finland (KELA) and Statistics Finland. A comparison against a local diabetes register of the Helsinki metropolitan area has demonstrated good coverage of diabetic patients in the nationwide database [6]. Follow-up data were obtained from the registers using the Finnish personal identification codes. In the present study, those who were not permanent residents of Finland or had gestational diabetes only were excluded from the analyses, and the type of diabetes was determined according to the study criteria. The Finnish Cancer Registry was launched in 1953 and is a national database of all cancer cases in Finland [7]. The National Board of Health established compulsory reporting of all diagnosed or suspected cases of cancer to the registry in 1961. The completeness of solid tumour registrations was reported to be over 99% in 1994 [8] and 96% of registered diagnoses were valid between 1985 and 1997 [9]. The period of follow-up in Finland was from 1972 until 2010.
Permissions to use the data for the study were received from all respective competent register authorities. The linkages were performed by competent authorities and the research group received anonymized data. Ethical approval for the study for received from the Research Ethics Committee of the Helsinki and Uusimaa hospital district.

Sweden
Launched as a quality assurance register in 1996, the Swedish National Diabetes Register includes patient demographic and clinical data such as treatments, complications and risk factor levels [10]. The registry is populated by data uploaded via clinical record databases by trained physicians and nurses following patient appointments at hospital outpatient clinics or primary health care centers. To ascertain the presence of and the date of diabetes diagnosis, both the Swedish National Diabetes register and the National Patient Register were searched for earliest mention of diabetes. For patients with two distinct dates of first diabetes mention, the earliest date was used as the date of diagnosis. The Swedish Cancer Register was established in 1958 and includes data on every primary malignancy, certain benign tumours and precancerous lesions in Sweden. Every clinician, pathologist and cytologist in Sweden must notify the National Board of Health and Welfare of each person who has been diagnosed with a new primary malignancy. In a comparison with the National Patient Register which comprises of information on all in-hospital care and outpatient specialist care in Sweden, the rate of non-reporting to the National Cancer Registry was estimated to be 3.7% in 1998 [11]. Data linkage between registries was achieved using the personal identification number, a unique patient number assigned to each Swedish citizen. For this study, patients diagnosed with type 1 diabetes between 1987 and 2011 were included.
This study was approved by the Central Ethical Review Board in Gothenburg.

Scotland
To determine patient diabetes status, the use of ICD-9 codes 250 and ICD-10 codes E10-E14 to identify diabetes diagnoses in Scottish General Acute/Inpatient (SMR-01) dataset was used because the national diabetes register was not complete until 2004. This dataset was established in 1960 and captures hospital admission data, including the primary medical condition which led to hospital admission and up to five other co-existing conditions, from all hospitals in Scotland [12]. A validation study using data from the national diabetes register after 2004 estimated that the SMR-01 captures approximately 72% of diabetes diagnoses in people aged below 30 [13]. The Scottish Cancer registry (Scottish Morbidity Record SMR-06) was established in 1958 and has been managed by the Information Services Division (ISD) of National Services Scotland since 1997 [14]. SMR-06 captures cancer registration data for the entire Scottish population, with approximately 47,000 cancer registrations made annually. Data on cancer diagnosis dates and cancer sites occurring between 1995 and 2011 were retrieved from SMR-06. To ascertain date and cause of death, data from the General Register Office of Scotland death records (SMR99) were retrieved. Data linkage between these registers was achieved using the Community Health Index, a unique patient identification number and probabilistic linkage. The study period for Scotland was from 1995 until 2011.
Approval for the creation and analysis of the Scottish linked dataset containing no personal identifying information was obtained from the Scottish Care Information -Diabetes Collaboration (SCI-DC) steering committee, the Scottish multi-centre research ethics committee (reference number 11/AL/0225), the Privacy Advisory Committee of National Health Service National Services Scotland and Caldicott guardians.

Australia
The National Diabetes Services Scheme was established in 1987 and currently captures data for approximately 80 to 90% of all people with a diagnosis of diabetes in Australia [15]. This registry was formed to provide subsidized diabetes products and information to people with diabetes in Australia. Registration to NDSS is free and is certified by medical practitioners. Prior to 2000, completeness of the diagnosis date variable was low and therefore a number of persons diagnosed with diabetes aged under 40 years during this time could not be included in these analyses [16].
We included type 1 diabetes who were registered on the NDSS in the years 2000 to 2008 (including all those registered before 2000 still alive on January 1 2000). The year 2000 was chosen as the start date as this time followed a unification of state-based registries as well as an improvement in data quality, including better capture of diabetes diagnosis date.
Type 1 diabetes status was assigned to registrants who satisfied all of the following conditions: • recorded as type 1 on the NDSS registry • diagnosed before 40 years of age • were taking insulin Registration date was used as a proxy for diagnosis date, as a large proportion of type 1 registrants (54%) had missing date of diagnosis, many of whom registered in the early years of the operation of the NDSS and had had diabetes for a number of years. We chose 40 years as the cut-off age to minimize the number of people with type 1 diabetes that we would miss, without misclassifying significant numbers of people with type 2 as type 1 [17]. Additionally, registrants who were recorded as type 2 on the registry, but diagnosed before age 30 and were taking insulin within 1 year of diagnosis date were reclassified as having type 1 diabetes.
The Australian Cancer Database was launched in 1982 and is a register of all primary, malignant cancers in Australia, where notification of cancer diagnoses is mandatory. The registry currently receives notifications on cancer diagnoses from several sources including hospitals, pathology laboratories, radiotherapy centers and death registries. Data linkage of these registries was completed by the Australian Institute of Health and Welfare using the Fellegi and Sunter framework [18]. This framework uses names first, second, third), sex and date of birth to conduct the linkage. Incident cases of cancer occurring between 1999 and 2009 retrieved from the website of the Australian Institute of Health and Welfare, at https://www.aihw.gov.au/acim-books/ (accessed March 2014); we used this slightly longer period for the population rates to get more stable estimates of the population rates in the period 2000-2008.
This study was approved by the Alfred Health Human Ethics Committee and the Australian Institute of Health and Welfare Ethics Committee. Table 1: Person-years and number of cancer cases in T1D patients (0-85 years) by sex, country and site. "All sites" comprise more that the single sites together, as other and unspecified sites are included.