www red DiabetologNytt
People with type 1 diabetes have reduced life expectancy (LE) compared with the general population. Our aim is to quantify mortality changes from 2002 to 2011 in people with type 1 diabetes in Sweden.
This study uses health records from the Swedish National Diabetes Register (NDR) linked with death records. Abridged period life tables for those with type 1 diabetes aged 20 years and older were derived for 2002–06 and 2007–11 using Chiang’s method. Cox proportional hazard models were used to assess trends in overall and cause-specific mortality.
There were 27,841 persons aged 20 years and older identified in the NDR as living with type 1 diabetes between 2002 and 2011, contributing 194,685 person-years of follow- up and 2,018 deaths. For men with type 1 diabetes, the remaining LE at age 20 increased significantly from 47.7 (95% CI 46.6, 48.9) in 2002–06 to 49.7 years (95% CI 48.9, 50.6) in 2007–11. For women with type 1 diabetes there was no significant change, with an LE at age 20 of 51.7 years (95% CI 50.3, 53.2) in 2002–06 and 51.9 years (95% CI 50.9, 52.9) in 2007–11. Cardiovascular mortality significantly reduced, with a per year HR of 0.947 (95% CI 0.917, 0.978) for men and 0.952 (95% CI 0.916, 0.989) for women.
From 2002–06 to 2007–11 the LE at age 20 of Swedes with type 1 diabetes increased by approximately 2 years for men but minimally for women. These recent gains have been driven by reduced cardiovascular mortality.
It has long been established that people with type 1 diabetes have higher mortality and reduced life expectancy (LE) com- pared with the general population [1, 2]. Having accurate LE estimates allows us to identify gaps between populations in order to identify areas for potential improvement and quantify improvements in healthcare over time [3, 4]. LE estimates feed into life insurance premiums and allow insurance com- panies to quantify and manage risks [5, 6]. The informationcan also be used by people with diabetes to make better plan- ning decisions about matters such as retirement .
Previous studies have estimated the impact of type 1 dia- betes on mortality and LE and compared these with variables in the general population [1, 8–14]. There is some evidence that survival of people with type 1 diabetes relative to the general population has improved since the 1940s  and has continued to improve in recent years . One Australian study comparing people with type 1 diabetes with the general population estimated that the standardised mortality ratio (SMR) decreased from 4.20 in 1997 to 3.08 in 2010 . Similarly, LE has improved over time . However, the gap still remains. A recent study in Scotland found that from 20 years of age, men and women with type 1 diabetes lose about 11 and 13 years of LE, respectively, compared with the general population . Renal and cardiovascular disease (CVD) have been identified as comorbidities that contribute to excess mortality, especially after the first 10 years from type 1 diabetes onset [12, 14, 18–21]. There is some evidence to suggest that improvements in diabetes care have delayed the progression of renal disease .
The current paper addresses whether mortality and LE have changed for men and women with type 1 diabetes in Sweden from 2002 to 2011, and examine how these changes compare with changes in the general population. In addition, we also explore whether any secular trends in mortality over this pe- riod can be explained by improvements in risk factors associ- ated with diabetes-related complications and further investi- gate mortality by cause to examine where recent gains have been made.
The estimated mortality rates of those with type 1 diabetes at different ages were comparable but slightly lower than those found in other studies . From 2002–06 to 2007–11 the LE of those with type 1 diabetes in Sweden increased for men by about 2 years, with no comparable evidence of any increase for women. However, this has not significantly closed the LE gap with the general population for men in absolute terms as a sim- ilar improvement was seen in the general population. Given that the mortality trend for men disappeared when we controlled for risk factors, the secular trend for men could potentially be ex- plained by changes in risk factors. When mortality was broken down into CVD, renal disease and other causes, we found that mortality relating to CVD significantly decreased from 2002–10 for both men and women. The improvement in CVD mortality coincided with the large increase in the proportion of the popu- lation with type 1 diabetes who reported being on lipid-lowering medication and the associated decrease in cholesterol over this period. However, similar relative improvements in the general Swedish population for CVD were also observed, which sug- gests a similar uptake in lipid-lowering medication in the general population . Also, reductions in smoking rates for the population with type 1 diabetes were also likely to have contributed to the lower CVD mortality. For Swedes with type 1 diabetes at the age of 20, the LE gap vs the general population was about 10–11 and 11–12 years for men and women, respec- tively. Interestingly, the secular trend in mortality was no longer significant for men when we controlled for metabolic risk factors, many of which have a significant impact on mortality (e.g. the association between triacylglycerol level and mortality for both men and women) . The seemingly counterintuitive results for BMI may be because a high BMI is a marker for good historic glycaemic control—in the DCCT study the intensive- treatment group had significantly higher BMI —or because the positive effects of greater lean mass in overweight and obese older people counterbalance the negative effects of greater fat mass on mortality . In addition, low systolic BP may be a marker for underlying poor health rather than a cause of mor- tality, particularly in people with heart failure .
The Swedish LE results for 2007–11 can be compared with evidence from Scottish people with type 1 diabetes (2008–10), where the remaining LE at age 20 for men was estimated to be 46.2 years (95% CI 45.3, 47.3) and for women 48.1 years (95% CI 46.9, 49.3) . Thus, 20 year olds with type 1 diabetes in Sweden have, on average, an LE 3.5 (95% CI 2.2, 4.8) and 3.8 years (95% CI 2.2, 5.2) longer than their Scottish counterparts for men and women, respectively. The Swedish type 1 diabetes LE gaps with the general population at age 20 are smaller than the Scottish gaps by 0.8 (95% CI −0.5, 2.2) and 0.9 (95% CI −0.6, 2.4) years for men and women, respectively, though these are not significant . The SMRs estimated in this study are of a similar magnitude to those found in a cohort covering a similar time span in Denmark , but were significantly lower than the SMRs found for men (3.8) and women (5.8) in a recent meta- analysis . The ratio of female to male SMRs was lower in Sweden for the period 2002–06 (1.12), but higher for the period 2007–11 (1.46) than the ratio in a recent meta-analysis (1.37) . However, care needs to be taken when making conclusions by comparing SMRs across populations with differences in un- derlying mortality risk because a larger relative difference in mortality may still be a smaller difference in absolute terms [42, 43]. This also applies to the current paper where gaps in LE are reported in absolute terms while the trends in the Cox regressions examine mortality differences in relative terms.
Compared with previous studies that have estimated LE in people with type 1 diabetes [5, 17], the strengths of this study are the large representative sample, the long length of follow- up, measurement of clinical risk factors and completeness of data. This is attributable to the NDR data, a large register that has been linked to hospital, prescription and death data. The NDR includes the vast majority of the Swedish type 1 diabetes population; 91% of 18–34 year olds identified with type 1 diabetes from a 2009 prescribing registry data could be matched to those in the NDR . This makes the consistent estimation of life tables for this population possible. However, participation in the NDR is not compulsory and our analysis
requires patients to make at least one clinic visit in or after 2002; hence our analysis may have excluded a small propor- tion of the older and sicker patients in the earlier years. The NDR tracks patients using their personal identification num- ber. However, for patients who emigrate it is not possible to know whether or when they died, or information that would allow them to be censored. Those alive and without at least one insulin prescription filled per year after 2005 were exclud- ed from the analysis, and this should have excluded all those with type 1 diabetes who emigrated. Given the short-term nature of the data available, the LE estimates are from period life tables rather than cohort life tables. This means that recent improvements in care that take time to translate into reduced mortality may not show up in such an analysis. In addition, there may be misclassified causes of death , with renal- disease-related deaths misclassified as CVD or vice versa . The Swedish renal mortality rates reported here are lower than those previously seen in a younger Japanese and US cohort where an expert committee based the cause of death classification on information from the attending physician or death certificates [14, 46–48]. However, the rates of end stage renal disease (ESRD) for those with type 1 diabetes in Sweden have previously been found to be low .
There is still some way to go in terms of improvement in care for those with type 1 diabetes in order to close the gap with the general population. A significant proportion have elevated HbA1c levels and a recent paper based on the Swedish NDR highlighted the stark differences in mortality for those with well-controlled vs poorly controlled HbA1c . In addition, with 10% of men and 13% of women still reported as current smokers in 2011, additional smoking cessation programmes could generate further improvements. While there have been large increases in the use of lipid-lowering medication, further expansion could generate additional gains given this population’s high underlying CVD risk. Future research might also provide individual specific LE estimates based on an individual’s characteristics in terms of their age at diagnosis, and risk factor and comorbidity profiles. This would provide useful information for an individual and allow them to better grasp the likely benefits of improving their overall risk.