WHAT IS ALREADY KNOWN ON THIS TOPIC

WHAT THIS STUDY ADDS

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

The study cohort was included from the Swedish Apolipoprotein Mortality Risk (AMORIS) cohort,13 consisting of 812 073 individuals, primarily from the Stockholm area, referred for laboratory testing from either occupational healthcare or primary care during the period 1985–1996. From this cohort, we included subjects confirmed to be fasting overnight prior to blood sampling, and with measured glucose, total cholesterol and triglycerides at the same examination (index examination). Subjects with diagnosed aortic valve disease (online supplemental table 1) or age below 18 years at index examination were excluded. The final study population consisted of 324 449 subjects (figure 1).

Information on comorbidities was obtained through linkage to the National Patient Register,14 the National Cancer Register15 and the National Cause of Death Register.16 International Classification of Diseases (ICD) codes used to identify comorbidities are presented in online supplemental table 2. Information on level of education and socioeconomic status as blue-collar or white-collar worker was obtained from the national population and housing census closest in time to the index examination up to 1990, after which socioeconomic data were obtained from the longitudinal integration database for health insurance and labour market studies. Record linkage was possible using each individual’s unique Swedish personal identification number.17

Information on analysed biomarkers was obtained from the index examination for all subjects. All blood samples were analysed in fresh blood at the CALAB medical laboratory in Stockholm, with well-documented and consistently applied analytical methods. For detailed information regarding analytical laboratory methods, see online supplemental material.

Subjects were followed for incident aortic stenosis or aortic valve disease through the National Patient Register and the National Cause of Death Register using registered ICD codes. With the implementation of the ICD-10 system in 1997, aortic stenosis gained a specific ICD code. Prior to this, the ICD code for aortic valve disease was used to identify outcomes (online supplemental table 3). ICD codes were registered by physicians from inpatient or emergency care, and from 2001, also from specialised outpatient care. Mean follow-up time was 25.9 years. Subjects were followed until registered diagnosis of aortic valve disease (1985–1996) or aortic stenosis (1997–2020), death from any cause, migration or end of follow-up on 31 December 2020.

Outcome was defined as registered diagnosis of aortic valve disease in the ICD-8 or ICD-9 system, or as registered diagnosis of aortic stenosis in the ICD-10 system. Exposure was defined based on fasting glucose at the index examination, in subsequent analysis categorised as either low (<3.9 mmol/L), normal (3.9–6.0 mmol/L), impaired fasting glucose (IFG) (6.1–6.9 mmol/L) or high (≥7.0 mmol/L), with subjects with diagnosis of diabetes mellitus categorised separately. In addition, the defined glucose levels for IFG according to the American Diabetes Association (ADA)18 (5.6–6.9 mmol/L) was used for separate analysis.

Dysglycaemia was defined as having a fasting glucose level at or above 6.1 mmol/L or a diagnosis of diabetes. Body mass index (BMI) calculated as kg/m² was available from the index or earlier examinations of the AMORIS cohort, the Swedish National Diabetes Register,19 the Swedish Medical Birth Register20 or research cohorts previously linked to the AMORIS cohort.13 Registered BMI within a timeframe of 5 years prior to or after the index examination was used, and if several measurements were available, we used that closest in time to index examination.

Baseline characteristics were described using frequencies and percentage for categorical variables and mean and SD for continuous variables. Cox proportional regression analysis was used to estimate HRs with 95% CIs. The proportional hazards assumption for all covariates was assessed visually using log-log survival curves. Adjustments were made for age and sex, with further adjustments for total cholesterol, triglycerides, socioeconomic status, education and diagnosed hypertension and chronic kidney disease. Age-stratified analysis (age <50 years, 50–69 years and ≥70 years) was performed. Sensitivity analysis was performed in subgroups with information on BMI and apolipoprotein B (apoB)/apolipoprotein A-1 (apoA-1) ratio, respectively. To examine the association with fasting glucose as a continuous variable, HRs for fasting glucose were estimated using restricted cubic spline analysis with 95% CIs, using a glucose level of 4.2 mmol/L as the reference. Five knots were used with the first set at fasting glucose 4.0 mmol/L, the third at 4.8 mmol/L and the fifth at 6.2 mmol/L. A nested case-control study was performed to explore to what extent fasting glucose levels differed between those with incident aortic stenosis versus matched controls at different time points prior to the diagnosis. Cases were defined as all subjects with a new diagnosis of aortic valve disease or aortic stenosis during follow-up. Five controls per case was randomly selected through incidence density sampling and matched to cases by age and sex. Statistical analysis was performed using STATA V.17.0 (StataCorp, College Station, Texas, USA).

Table 1

Baseline characteristics

During a mean follow-up of 25.9 years, 8523 events of aortic stenosis occurred, resulting in a cumulative incidence of 2.6%. Of these, 897 (10.5%) were diagnosed with aortic valve disease before the implementation of ICD-10 in 1997. An association with incident aortic stenosis was seen in all groups with dysglycaemia with an HR (95% CI) of 1.36 (1.24 to 1.5), 1.79 (1.60 to 1.99) and 2.21 (1.80 to 2.73) for subjects with IFG, high glucose and diagnosed diabetes, respectively (table 2).

Subjects who developed aortic stenosis during follow-up had a higher mean age at baseline (54.4 vs 44.6) and a higher proportion of males, as well as higher fasting glucose, lipid levels, apoB/apoA-1 ratio and CRP.

Cubic spline analysis showed a continuously stronger association with aortic stenosis with increasing glucose levels, starting already below the diagnostic threshold for IFG (6.1 mmol/L) (figure 2).

When comparing ADA criteria for IFG with that of the WHO criteria, further analysis with IFG categorised according to the definition by the ADA (5.6–6.9 mmol/L) resulted in an additional 23 374 subjects categorised as having IFG (n=33 727). For this group, the observed HR showed a similar pattern to those categorised according to the WHO criteria, although slightly lower (HR 1.24 vs 1.36) (online supplemental table 4).

In the nested case-control analysis, higher levels of glucose in cases compared with controls were consistently present >30 years prior to the diagnosis of aortic stenosis and became more pronounced closer in time to the diagnosis (online supplemental figure 1).

When exploring the association by age groups, the association were strongest among those at younger age at inclusion (online supplemental table 5). For the oldest age group (≥70 years at inclusion), the association showed a similar pattern to the overall analysis, but the association with IFG no longer remained statistically significant.

Subgroup analysis among subjects with available apoB/apoA-1 ratio (n=68 435) showed an increased association with aortic stenosis among all groups with elevated glucose, with the highest association observed in those with known diabetes. Adjustment for the apoB/apoA-1 ratio did not affect this association among those with diabetes and only slightly attenuated the association for those with IFG and high glucose (online supplemental table 6).

Further subgroup analysis of subjects with available data on BMI (n=53 889, baseline characteristics in online supplemental table 7) showed a strong association with aortic stenosis for those with high glucose levels or diagnosed diabetes. These associations appeared to be somewhat attenuated after adjusting for BMI (HR 1.72 and 2.59 for high glucose and diabetes, respectively). For IFG, the association was weaker and the point estimate after adjustment for BMI suggested no association with aortic stenosis, although CIs were wide and overlapping (online supplemental table 8).

Elevated low-density lipoprotein (LDL) cholesterol, decreased levels of apoA-1 and elevated levels of apoB are associated with increased risk of aortic stenosis.21 22 Statin treatment has not been shown to be effective in reducing this risk.23 24 However, a secondary analysis of the SEAS study, evaluating simvastatin and ezetimibe combined, showed a 60% reduced risk for valve replacement among patients with mild aortic stenosis and LDL >4 mmol/L.25 Additionally, there are ongoing trials studying the effect of other lipid-lowering treatments on the progression of aortic stenosis.26 In our study, information on apolipoprotein levels was available in a subgroup (n=68 435), and adjusting for apoB/apoA-1 ratio on top of total cholesterol and triglycerides had essentially no effect on the association between dysglycaemia and aortic stenosis.

The development of degenerative aortic stenosis is initiated through endothelial damage and successive lipid infiltration and inflammation.5 This in turn causes the localised deposition of calcium crystals, which maintains an inflammatory response and progresses the disease into a second fibrotic phase. At this stage, the disease progression is driven by fibrosis and calcification, causing osteogenic processes resulting in pronounced calcification and loss of function of the aortic valve.6 Several mechanisms through which diabetes may drive progression to aortic stenosis has been suggested. When examined ex vivo, aortic valves from patients with diabetes have increased levels of advanced glycated end-products, in turn associated with a smaller aortic valve area, indicating more pronounced stenosis.27 Furthermore, ex vivo studies have shown higher expression of nuclear factor kappa B and bone morphogenetic protein 2, which are involved in the progression of aortic stenosis, in aortic valves from individuals with diabetes.28

Few studies have examined the association of glucose levels with incident aortic stenosis. A large observational study investigating the risk for valvular heart disease among patients with type 2 diabetes found that although diabetes increased the risk for aortic stenosis, higher haemoglobn A1c (HbA1c) did not affect this risk.9 It has been suggested that metabolic risk factors mainly influence the initial phase in the establishment of aortic stenosis, and less in the continued progression of the disease.11 The initial establishment of degenerative aortic stenosis is characterised by endothelial damage and inflammation.26 Hyperglycaemia and prediabetes are associated with increased levels of inflammatory markers and cause an upregulation of pro-inflammatory cellular pathways, leading to increased oxidative stress.29 After the establishment of initial calcification however, continued disease progression becomes, in part, self-propagating.6 Thus, elevated glucose levels may influence the establishment of aortic stenosis, and it is possible that this occurs already at lower glucose levels and before the diagnosis of diabetes. This would be consistent with our findings of an association already at prediabetic levels of glucose. Furthermore, the nested case-control study showed that, decades prior to diagnosis, subjects with incident aortic stenosis had higher fasting glucose levels than controls.

Information on BMI was only available for a subgroup (n=53 889), in which IFG was not shown to be associated with aortic stenosis. This subgroup did differ from the total study cohort in some respects, however, particularly in having a lower prevalence of cardiovascular comorbidity and a higher proportion referred via occupational healthcare (80% vs 56% in the total cohort), suggesting that subjects with measured BMI may have been healthier at baseline.

Adjusting for BMI attenuated the risk for aortic stenosis among those with high glucose levels and diagnosed diabetes. Due to the small size of this subgroup, these results must be interpreted with caution, but this suggests that BMI remains a potential confounder. A recent study on Swedish patients with diabetes did indeed find BMI to be associated with aortic stenosis.9

With the expanding role of novel glucose-lowering drugs in the treatment of other cardiovascular diseases, there are opportunities to study whether these may also have a preventive effect on aortic stenosis. Glucagon-like peptide-1 receptor agonists, which lower blood glucose and facilitate weight loss, may be of particular interest. Given the biphasic pathophysiology of aortic stenosis, which involves initial inflammation and accumulation of lipoproteins, it is possible that treatment of metabolic risk factors may have a preventive effect, particularly in the early stages of the disease.

The main strengths of this study are its large size and long follow-up of this population-based cohort, along with a wide array of biochemical analyses performed at baseline. All analyses were conducted on fresh blood samples in the same laboratory, ensuring consistency and comparability of the biomarker information. Furthermore, most subjects were referred for testing through routine health check-ups via occupational healthcare, and the overall prevalence of comorbidities was low, with no subjects hospitalised at the time of index health examination. Linkage to several high-quality national registries enabled precise follow-up and provided additional information, such as prevalent comorbidities and socioeconomic status.

The limitations of this study are mainly due to the registry-based methodology, which resulted in a lack of information on lifestyle factors, including smoking status. Additionally, we did not have information on blood pressure or pharmacological therapy at baseline. Therefore, information on hypertension at baseline was defined as a registered diagnosis of hypertension in the National Patient Register. Although the ICD system has high validity in the National Patient Register, it has previously been shown to have limited coverage for hypertension specifically.14 30 Thus, the prevalence of hypertension in our cohort has likely been substantially underestimated and remains a potential confounder. Furthermore, HbA1c was not included, as it was not routinely analysed at the time of inclusion and was largely unavailable. Since HbA1c reflects prolonged exposure to elevated glucose levels, it is possible that elevated HbA1c could show a stronger association with incident aortic stenosis. It is possible that we would have identified more subjects as having dysglycaemia if HbA1c had been available, meaning the association based on fasting glucose levels could be underestimated. As HbA1c has become widely used in clinical practice to detect dysglycaemia, this is an important area for further studies.

In a Swedish setting, the diagnosis of aortic stenosis is based on echocardiographic findings, with the examination typically referred either due to the auscultation of a heart murmur or as part of an evaluation for symptoms such as breathlessness, chest pain, syncope or signs of congestive heart failure. Echocardiography is not performed without prior suspicion of cardiovascular disease, and we do not believe that subjects with elevated fasting glucose received echocardiography more frequently than those with normal glucose levels. However, it is possible that subjects with a diagnosis of diabetes were monitored more regularly, which could contribute to a higher incidence of diagnosed aortic stenosis in this group.

Not applicable.

This study was approved by the Regional Ethical Review Board in Stockholm (Reg. No. 2018/2401-31) and complies with the Declaration of Helsinki. Subjects gave informed consent to be included in the AMORIS cohort and subsequent research on this cohort.