Using this population-based register of people with diabetes in Scotland, we have reported contemporary rates and trends in type 2 diabetes incidence and mortality between 2004 and 2013 and examined the effect of age, sex and socioeconomic deprivation on these estimates.
Type 2 diabetes incidence was found to be relatively stable after 2004 while mortality rates declined. The number of new cases of type 2 diabetes exceeded the number of diabetes deaths. Important differences in trends in incidence and mortality rates were observed by age, sex and socioeconomic status. Despite improvements in absolute mortality rates, type 2 diabetes continues to confer an excess risk of death compared with the non-diabetic population.
Relation to other studies
Declining or stable type 2 diabetes incidence in the mid-2000s has also been observed in other populations [
5,
11‐
13]. A register-based study in Denmark, which was unable to distinguish by diabetes type, reported increasing diabetes incidence between 1995 and 2004 but declining incidence between 2004 and 2007 [
11]. In Sweden, the incidence of pharmacologically treated diabetes decreased modestly between 2005 and 2013 [
13]. The National Health Interview Survey in the USA identified that diabetes incidence doubled between 1990 and 2008 before plateauing between 2008 and 2012 [
12]. Type 2 diabetes incidence stabilised after 2005 in the UK according to data from The Health Improvement Network (THIN) database [
5].
In contrast, another UK-based study conducted using the Clinical Practice Research Datalink database reported that standardised incidence ratios of diabetes increased steadily from 100 (referent period) between 1991 and 1995 to 275 (273–276) between 2006 and 2010 [
14]. Increasing age-standardised type 2 diabetes incidence among men participating in the British Regional Heart Study was observed between 1986 and 2007 [
15]. These conflicting findings may reflect differences in diabetes definition, dates of comparison, methods of calculating disease incidence and underlying population characteristics.
The largest fall in type 2 diabetes incidence in Scotland took place between 2004 and 2005. The introduction of the Quality and Outcomes Framework (QOF) in the UK in 2004, which provided incentives for general practitioners to have a diabetes register, may have contributed to this finding.
A proposed explanation for declining or stable diabetes incidence after 2005 is a reduction in the pool of undiagnosed diabetes through the intensification of diagnostic activities during the last decade [
11,
16]. In 1998, the World Health Organization altered the diagnostic criteria by lowering the threshold for diagnosing diabetes [
17], a factor which may have contributed to the marked increases in diabetes incidence in the late 1990s and early 2000s. For example, type 2 diabetes incidence doubled between 1993 and 2004 in Tayside, Scotland. This increase may partly reflect improved diagnosis of diabetes in addition to the change in diagnostic criteria [
18]. This proposed explanation is also supported by several studies which have reported declining numbers of undiagnosed diabetes cases during the last 20 years [
19‐
21]. Data from the German Health Interview and Examination Surveys indicate that the proportion of undiagnosed diabetes decreased from 40% in 1997 to 22% in 2010 [
19]. Furthermore, the proportion of diabetes cases which were undiagnosed in Scotland declined from 17% in 2010 to 12% in 2013 according to estimates from the Association of Public Health Observatories diabetes prevalence model, although it is not known if the proportion was higher in the early 2000s [
22]. Our finding of differences in incidence trends by age may reflect different patterns of opportunistic screening and obesity trends.
Stable prevalence of adulthood obesity, an established risk factor for type 2 diabetes, is also likely to have influenced our finding of stable diabetes incidence [
23]. Estimates from the Scottish Health Surveys indicate that the proportion of adults who were obese increased from 24.2% in 2003 to 27.1% in 2009 but remained constant thereafter. Moreover, trends in age-specific obesity prevalence mirror trends in age-specific diabetes incidence [
23]. For example, data from the Scottish Health Surveys suggest that the proportion of obese men aged 35–44 years peaked in 2009 at 31.9% before declining to 24.7% by 2013; a similar pattern was observed in men aged 45–54 years. This trend coincides with a peak in diabetes incidence in 2009 before declining thereafter among men aged below 65 years.
The disparity in incidence rates between men and women has been observed in previous studies [
11,
16]. One explanation for this is the higher risk associated with male body fat distribution and greater insulin resistance among men relative to women [
24]. Findings from a study conducted using the SCI-Diabetes dataset support this notion with men developing type 2 diabetes at lower BMIs than women of a similar age, with particularly marked differences at younger ages [
25]. Deprivation was also strongly associated with incidence of type 2 diabetes and socioeconomic inequalities in incidence of type 2 diabetes widened during the study period, particularly among women. Data from the Whitehall Study II suggest that health behaviours and BMI explained up to 45% of the differences in type 2 diabetes incidence by socioeconomic status [
26]. These findings underline the importance of targeting efforts to improve levels of modifiable risk factors in more deprived groups to achieve reductions in type 2 diabetes incidence and health inequalities.
Our finding of declining absolute mortality among people with type 2 diabetes in recent years corresponds with results from several studies conducted in developed countries, including Canada [
16], the UK [
27], Denmark [
11,
28], Australia [
29] and the USA [
30]. Increased testing may have led to earlier diagnosis of diabetes or diagnosis of people with different characteristics from those diagnosed in the early 2000s. However, while previous studies have reported declining values of glycated haemoglobin at diagnosis during this study period [
31], there was no evidence of a downward trend in glycated haemoglobin levels at diagnosis of type 2 diabetes by year using data from SCI-Diabetes (see ESM Table
2).
Improved management of hypertension, cholesterol and diabetes, as well as reductions in smoking, are also likely to have contributed to the observed declines in mortality rates. During the last decade, diabetes treatment guidelines have emphasised the need for intensive risk factor control [
32,
33], while the implementation of the QOF has contributed to improved diabetes care and monitoring. Indeed, the 2014 Scottish Diabetes Survey reported that the proportion of people with type 2 diabetes and a record of HbA
1c below 7.5% (58 mmol/mol), reflecting good diabetes control, increased from 50% to 61% between 2004 and 2013 [
2]. Similarly, the proportion of people with diabetes (not stratified by type 1 and 2) with a systolic blood pressure below 140 mmHg increased from 63% to 79% between 2004 and 2013 [
2].
However, if the improvements in mortality represent improved diabetes care or earlier diagnoses of type 2 diabetes then it might be expected that the SMRs in people with type 2 diabetes compared with people without type 2 diabetes would also have declined. We found no change in SMRs during the study period, with mortality rates remaining approximately 40% and 80% higher among men and women with type 2 diabetes, respectively, relative to people without diabetes. These findings contrast with earlier studies which have reported declining SMRs over time and a smaller overall relative influence of type 2 diabetes on mortality [
11,
27,
29,
30]. For example, in Australia, SMRs declined from 1.40 (95% CI: 1.36, 1.44) in 1997 to 1.21 (1.19, 1.23) in 2010 in men, and from 1.56 (1.51, 1.61) to 1.22 (1.19, 1.24) in women [
29]. Findings from a study using the THIN database reported declining age and sex adjusted SMRs between 1996 and 2009, falling from 2.14 (1.97, 2.32) to 1.65 (1.57, 1.72) [
27]. In a Swedish registry study, excess risk of mortality was 27% between 1998 and 2011 after adjusting for age and sex, decreasing to only 15% when additionally adjusting for country of birth, educational level and comorbidity [
34]. A time interaction was also reported in this study with type 2 diabetes conferring a 17% excess risk of mortality compared with controls matched for age, sex and county between 1998 and 2001, a figure which declined to 13% between 2005 and 2011. Explanations for lower excess risks of mortality among people with type 2 diabetes in these countries compared with Scotland include methodological differences between the studies such as additional adjustment for confounders, smaller differences in the health status between people with and without diabetes or differences in diabetes care.
Our findings of marked sex differences in SMRs associated with type 2 diabetes (See ESM Table
3) is in contrast to findings from earlier studies [
11,
27,
29,
30]. Unfortunately, only one of these studies reports absolute mortality rates by sex and diabetes status, thus making it difficult to determine whether the differences in SMRs are primarily driven by absolute mortality among women with type 2 diabetes or mortality among the non-diabetic population. Further investigation is required to identify explanations for higher SMRs associated with type 2 diabetes both in comparison to other settings and between women and men in Scotland.
Mortality rates were higher in the most deprived groups compared the least deprived groups. Similar patterns have been observed in several earlier studies [
3,
35‐
38]. We have shown that socioeconomic inequalities in mortality have persisted throughout the study period. The explanation for poorer survival among the most deprived groups is likely to include differences in risk factor prevalence and control, and healthcare provision and use. In the Scottish diabetes register, the mean age of people with newly diagnosed type 2 diabetes was 59.2 years among the most deprived decile compared with 63.1 years in the least deprived decile. This variance is probably related to differences in obesity prevalence by deprivation (62.4% and 51.5% of obese people in deprivation deciles 1 and 10, respectively). A previous study among people with type 2 diabetes in Scotland found the expected differences in smoking habits but no differences in cholesterol control, blood pressure or glycaemic control by socioeconomic status [
39].