Introduction
Global trends on the burden of diabetes and heart failure (HF) continue to increase, thus highlighting the need for more effective preventive strategies [
1,
2]. Type 2 diabetes (T2D) and HF individually confer considerable burden that are multiplied when these diseases co-exist, further decreasing patient’s quality of life and increasing healthcare costs [
1]. In Switzerland, 500,000 (6%) of the population is estimated to have T2D and about 30–40% of them have chronic kidney disease (CKD) [
3]. Meanwhile, cardiovascular diseases are main cause of mortality accounting for 27% of all deaths in 2020 [
4], likely preceded by heart failure.
Hospitalised patients with HF represent advanced stages of disease with poor prognosis: high risk of in-hospital mortality from 4 to 7%, rehospitalisation from 25 to 30%, and mortality from 7 to 11%, with shorter life expectancy [
2,
5]. While T2D is expected to worsen the prognosis of patients with HF, the evidence from observational studies has been inconsistent, with some studies showing adverse impact [
6,
7], and some other studies reporting no impact on mortality [
8,
9]. Moreover, these studies have mostly described short- and intermediate-term outcomes (i.e. up to 1-year) [
6‐
9]. Meanwhile, CKD is prevalent among patients with T2D, and HF patients with CKD show higher risks for mortality than those without CKD [
10]. Whether co-existence of both T2D and CKD in inpatients with HF would pose a higher burden on long-term mortality and life expectancy is not yet fully explored.
Studies that have explored the association between diabetes and mortality in HF patients did not estimate life expectancy of inpatients with HF, nor compare life expectancy between those with or without diabetes [
11]. Previous studies limited their findings to hazard ratio estimates (instead of absolute risk estimates), that limits bed-side interpretability [
7,
12,
13]. Similarly, some studies adjusted for presence of CKD, and but did not estimate the added mortality risk of CKD among inpatients with HF [
11‐
13].
Therefore, by using data of inpatients with HF at the largest tertiary cardiovascular referral hospital in Switzerland, we compared mortality outcomes among patients with or without T2D and estimated the differences in life expectancy between groups. Additionally, we explored whether co-existence of T2D and CKD was associated with worse prognosis.
Discussion
In this ambispective, observational clinical and civil registry of consecutive inpatients with HF in Switzerland’s largest tertiary cardiovascular referral center, we described the country’s largest cohort inpatients with HF and analysed long-term outcomes. We observed that overall, T2D was independently associated with 21% higher risk of all-cause mortality and shorter life expectancy among inpatients with HF, even after taking into account multiple risk factors and potential confounders such age, sex, CKD, EF, ASCVD, hypertension, dyslipidemia, atrial fibrillation, and COPD. Furthermore, presence of both T2D and CKD was associated with 84% higher risk of all-cause mortality and reduced life expectancy. In patients with HFpEF, T2D confers higher risk of all-cause mortality and shorter life expectancy. CKD confers significantly higher risk than T2D, and even higher when both T2D and CKD coexist. In patients with HFrEF, the higher risks conferred by T2D or CKD were similar, but higher when both coexist.
Our study population is consistent with a large multinational study describing a contemporary population of patients with heart failure, where ASCVD such as ischemic heart disease and CKD are found to be common comorbidities [
16]. Moreover, our results are in line with previous studies that have compared clinical characteristics and long-term outcomes of patients hospitalized for heart failure with or without diabetes [
11,
12,
17]. In a meta-analysis of observational studies, diabetes was associated with higher risk of all-cause, long-term mortality based on pooled estimates from acute and chronic HF registries (HR 1.13, 95% CI 1.05 to 1.22,
I2 = 85.3, and HR 1.44, 95% CI 1.36 to 1.52,
I2 = 56.1, respectively) with longer follow-up periods up to 15 years [
11]. A randomized controlled trial with 5 to 8 years of follow-up also showed diabetes as an independent risk factor for all-cause mortality [
12]. Moreover, our study also is also in line with a large observational study where CKD was associated with further increases in mortality risks among patients with HF and T2D [
17]. Nevertheless, our study extends previous findings by exploring the influence of CKD and T2D in HF patients, and extrapolating association estimates to life expectancy estimates.
In our analysis that stratified patients according to EF category, we did not observe a statistically significant interaction between T2D and EF category for all-cause mortality. This is in line with findings from the CHARM program, in that the magnitude of all-cause mortality did not differ between EF categories [
13]. They also did not observe an interaction of T2D and EF with cardiovascular mortality as the outcome [
13]. Potential interaction effects may not have been observed in our population since treatment significantly improves prognosis in HFrEF but only modestly in HFpEF,
We observed that the association of T2D on HF could be time-dependent, with similar risk of mortality in the acute term (< 30 days follow-up), but an increasing risk of mortality with longer follow-up even after accounting for age. This could partly be explained by the delayed effects of diabetes on HF, and the episode of hospitalization accelerated the longstanding maladaptive alterations, structural changes, cardiovascular dysfunction that T2D already imparts before hospitalization [
18]. Moreover, our results may suggest a paradoxical benefit (i.e. HR < 1.0) of CKD in the acute term (< 30 days) in HF patients and deleterious effects afterwards. This could in part be explained by underdiagnosis of CKD in this population and hence undertreatment, whereas those who have been diagnosed early might have received appropriate treatment that may improve survival early during hospital admission. This hypothesis could be further strengthened by recent evidence showing that in general, many cases of kidney disease may be missed when using medical record data [
19]. Thus, the possibility of misclassification of undiagnosed CKD with worse acute prognosis in the non-CKD group could be plausible. Investigation of these maybe warranted if acute survival is of particular interest.
Overall, we observed that T2D and CKD were associated with higher risk of mortality among HF inpatients in the long term (i.e. beyond acute period). We extended evidence by translating relative risks into absolute measures more interpretable to the patients, clinicians, and policymakers. Inpatients with heart failure and T2D had shorter survival than those without T2D, those with CKD had shorter survival than those with T2D whilst those with both T2D and CKD had shorter survival than those with either T2D or CKD alone. These estimates constitute a substantial life expectancy reduction, considering that our patient population is relatively older and represents patient populations that already have a very high mortality risk. Moreover, healthcare costs at this age and disease severity are expected to increase, and particularly among patients with CKD who incur much higher hospital care costs, this could mean even much greater societal costs [
16].
The observed association of diabetes on long-term survival among HF inpatients could be explained by ischemic and non-ischemic mechanisms [
1,
18]: insulin resistance and hyperglycemia worsen dyslipidemia, accelerate atherosclerosis, and induce inflammation, promoting leucocyte adhesion and coronary plaque formation, eventually leading to plaque rupture and coronary thrombosis [
1]. Hyperglycemia activates renin–angiotensin–aldosterone system (RAAS) and cytokines, which cause endothelial dysfunction and vasospasm [
18]. Diabetes also increases transforming growth factor-β (TGF-β) directly through gene upregulation or indirectly through RAAS activation, which in turn increase TGF-β [
20]. TGF-β triggers a cascade leading to formation of cardiac fibrosis, leading to structural changes and dysfunction [
21]. In a meta-analysis, diabetes was associated with higher degrees of cardiovascular magnetic resonance imaging-derived estimates myocardial fibrosis [
22], and higher degrees of myocardial fibrosis have been associated with all-cause mortality [
23]. In patients with CKD, kidneys can release hormones and inflammatory cytokines that influence vascular tone, and hemodynamic alterations further affect the failing heart [
24]. Moreover, heart failure induced renal hypoperfusion activates RAAS, sympathetic nervous system, and arginine vasopressin that leads to fluid retention, increased preload, and worsening heart failure [
25].
To our knowledge, this is the first study that reported relative (i.e. HR) and absolute estimates (i.e. months survival) on the association of T2D with life expectancy among inpatients with HF, increasing bedside interpretability. We posit that after hospital admission with advanced stages of disease, life expectancy is a more useful information that could improve risk communication from clinicans to patients. Another strength of our study relies on its large, real - world hospital setting of inpatients with HF patients in Switzerland, exceeding the size of some multinational registries of HF in Europe [
26‐
28]. This large sample size provided us with sufficient power even when time-stratified and sensitivity analyses were performed. Through the hospital’s data science center, we were able to obtain information on mortality events over a relatively long period, thus we were able to estimate long-term mortality and robust life expectancy estimates. Because we included all inpatients with HF whether or not HF was the primary reason for admission, our results are thus more generalizable to the inpatient population. We were able to account for the relative impact of CKD, on the association between T2D and mortality, instead of only adjusting or stratifying as most studies have done.
Our findings should be interpreted against several considerable limitations. First, the data we analysed were based on medical records at the hospital. Thus, we lacked information on routine subclinical measures that stratify risk in patients with T2D or CKD: glycosylated hemoglobin and estimated glomerular filtration rate, that could have provided important insights clinical risk stratification. Further, data on cause of death were not available, which precluded us from exploring whether our findings were driven by CV deaths. This could have allowed us to compare results with other similar studies. Second, our study lacked information on durations of T2D, HF, and other comorbidities such as CKD. Our reliance on ICD-10 codes to identify most T2D and CKD cases could also have led to misclassification of non-T2D and CKD cases if the ICD code was not encoded, hence in the non-T2D group some cases of T2D maybe present. This is a pattern that is in general observed when using medical record data [
19]. However, if a negative association between T2D and mortality exists, that would have led to an underestimation of the association in our study, especially in long-term. Third, in our sensitivity analyses, we excluded patients who died during the same episode of hospital admission, and this may have introduced selection bias. However, mortality rates between those included and excluded were comparable. Fourth, we were not able to account for smoking, alcohol intake, and other lifestyle factors. However, we were able to account for most risk factors for mortality and confounders in the association between diabetes and mortality among patients with HF. Many patients lacked BMI measurements; however, mortality rates did not differ between groups in those with or without BMI measurements, suggesting low likelihood for selection bias to have occurred. Fifth, since our study involved high risk patients in Switzerland, these findings may only be generalized to patients with advanced risk profiles and those in similar healthcare settings. Finally, in the time stratified analyses, group sizes among females were small hence underpowered, and could likely explain the reason why no significant differences in risks were found between diabetes groups across all periods.
In conclusion, we showed that T2D is associated with a significantly higher risk of all-cause mortality with significant reduction in life expectancy among patients with HF, even when clinical risk factors and potential confounders such as age, sex, hypertension, EF, CKD, ASCVD, and other comorbidities were taken into account. Among those with T2D, CKD was associated with further reduction in life expectancy.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.