Lifetime risk of cardiovascular-renal disease in type 2 diabetes: a population-based study in 473,399 individuals

We used electronic health records (EHR) across primary care (Clinical Practice Research Datalink, CPRD-Aurum), hospital admissions (Hospital Episodes Statistics, HES), and death registry (Office for National Statistics, ONS) with prospective recording and follow-up; linked using unique national healthcare identifiers (CPRD and NHS Digital) [16]. Over 99% of England’s population is general practice (GP)-registered [1]. CPRD is representative by socio-demography, ethnicity and mortality.

We included individuals ≥18 years of age who had a record of T2D diagnosis for the first time between 1 April 2007 and 31 October 2018 and were registered with a general practice during the same period with at least one year of data prior to index date (1 Apr 2007). We excluded individuals with a general practice record of type 1 diabetes or gestational diabetes prior to index date (Fig. 1). Follow-up was from day after index date until the earliest of study outcome, death, transfer out of practice, or study end date (31 October 2018). Baseline comorbidities (any diagnosis record), medications (within 12 months) and investigations (latest) were searched in prescribed drugs, hospital records and general practice records prior to index T2D diagnosis, consistent with our prior analysis [1]. Specifically, baseline CVRD status was identified in primary care using Snomed codes (MedcodeIDs) for diagnoses and symptoms, as well as using ICD10 codes from hospitalisations. Outcomes were identified using ICD10 codes for in-patient hospitalisations listed as a diagnosis at any position. A complete list of definitions is available in Additional file 1: Table S1 and Additional file 17: Table S2, respectively. Data were analysed based on observed cases only without imputing of missing values, while those without the presence of diagnostic codes—prescription records of interest were assumed to be disease free—have not taken the drug.

For lifetime risk analysis, a subset of individuals aged 45–99 years at index date were included. Baseline CVRD was defined by history of angina; myocardial infarction, MI; heart failure, HF; stroke/transient ischaemic attack, TIA; peripheral artery disease, PAD; chronic kidney disease, CKD; or prescription for nitrates. The CVRD-free population was the reference group, and those with only one co-morbid CVRD were also looked at. Thus, the study population was stratified into mutually exclusive categories based on baseline CVRD: (1) CVRD-free; and only (2) HF; (3) CKD; (4) MI; (5) stroke; or (6) PAD.

The population attributable risk analysis was conducted in the overall study population. The population were stratified into high (exposed) or low risk (unexposed) groups for each baseline risk factor, as well as an overall cardiovascular health indicator.

The primary outcome is major adverse renal cardiovascular events (MARCE). MARCE were defined as a composite of cardiovascular (CV) death, and hospitalisations due to CVRD: including HF, CKD, MI, stroke or PAD. Secondary outcomes are the individual components of MARCE. In sensitivity analyses, we assessed major adverse cardiovascular events (MACE), defined as MI, stroke or CV death. We used validated EHR phenotyping algorithms and tools for national structured data [16].

The study population included 473,399 individuals with incident T2D: mean age 63.5 years (SD 14.4), 45.7% female, median follow-up 6 years [IQR 3–9], and 59% (n=276,941) CVRD-free, i.e. no known history of MI, HF, stroke/TIA, PAD, AKI, CKD, or prescription for nitrates. Individuals with no history of CVRD were younger (mean 58.3 years, SD 13.7) than those with history of HF (mean 70.2, S.D. 13.4), CKD (mean 71.8, S.D. 12.2), MI (mean 67.3, S.D. 12.4), stroke (mean 69.3, S.D. 12.6) and PAD (68.4, S.D.11.1). There were significant differences by gender: 46.4% female for those without history of CVRD, compared with HF: 47.9%, CKD: 58.3%, MI: 29.9%, stroke: 45.5% and PAD: 36.4% (Table 1).

Individuals without a history of CVRD had the lowest rates of atrial fibrillation (2.3%) and former smoking (42.0%), and the highest rate of HbA1c levels > 10% (10.5%). Blood pressure, cholesterol and triglycerides were similar across individuals, regardless of prior CVRD. Individuals without a history of CVRD had the lowest use of aspirin (14.2%), statins (40.4%), anti-hypertensive therapy (43.1%) and CVD risk-lowering therapy (59.6%) in the 12 months prior to index T2D diagnosis. However, there were gaps in evidence-based therapy for all patient groups, e.g. among those with prior MI, stroke and PAD, only 64.2%, 68.5% and 69.5% were taking statin, and 50.2%, 45.1% and 49.8% were taking aspirin. Use of glucose-lowering therapies was highest in those with PAD and lowest in those without prior CVRD (Table 1).

In individuals free from CVRD at age 45 years, lifetime risk of MARCE was 81% in men and 79% in women. Because remaining life expectancy and competing risks of death are considered, lifetime risk decreases with increasing age but remained high with increasing age, e.g. 77% and 75% at age 75 years, and 73% and 68% at age 85 years in men and women respectively (Fig. 2). In individuals with HF, CKD, MI, stroke and PAD, lifetime risk of MARCE was very high: 97.1% (95%CI 97.1-100.0), 93.3% (93.3–96.3), 97.6% (97.6–100.0), 89.2% (89.1–92.1) and 90.8% (90.8–98.9) respectively, compared with 80.0% (79.7–80.4) in those without CVRD (Table 2).

Lifetime risk of HF was highest in those with MI (44.0%, 37.2–49.8) and PAD (35.2%, 30.0–38.4), compared with 29.2% (27.5–29.7) in those without CVRD. For CKD, lifetime risk was highest in those with HF (66.0%, 65.2–70.6), but similar across other categories; CVRD-free: 53.6% (52.6-54.1), MI: 53.2 (49.1–58.8), stroke: 54.5 (52.9–56.7) and PAD: 56.0 (53.8–61.4). Of the major renal and CVD events, the lifetime risk of PAD was lowest across patient groups; e.g. CVRD-free: 9.2% (8.1–9.5), HF: 12.8% (8.8–15.5), MI: 14.3% (9.1–17.9%).

Lifetime risk of MI and CV death varied by a history of pre-existing CVRD and were highest in those with HF and PAD. The lifetime risk of stroke was lowest in individuals with HF (18.7%, 13.2–21.7), compared with 20.2% (18.5–20.6) in individuals without history of CVRD (Fig. 2, Table 2). Conversely, lifetime risk of HF was lowest in those with stroke (28.8%, 25.5–30.5), compared with those without CVRD (29.2%, 27.5–29.7).

Out of the individual cardiovascular and renal comorbidities, ARE of MARCE was highest in T2D patients with HF at 0.748, i.e. in those with HF only, 75% of MARCE after index T2D can be attributed to HF. Nearly 5% of incident MARCE is attributable to CKD (PAR 0.046) and 2% (PAR 0.019) to stroke. The PAR is lower for HF compared to CKD due to fewer individuals with HF only (n=5310, 1.12%) compared to CKD (n=35,019, 7.40%), as PAR accounts for both strength of increased risk and prevalence of the risk factor. Among other modifiable risk factors, incidence of MARCE in severely obese T2D patients would decrease by 30% by modification to normal BMI. If the BMI of severely obese people is reduced to normal levels, 3% of MARCE in the T2D population could be prevented (PAR=0.031).

Smoking, poor HbA1c control (> 10%) and hypertension are risk factors for MARCE: odds ratio (95% CI) 1.15 (1.12–1.19), 1.96 (1.88–2.04) and 1.24 (1.21–1.27), respectively, after adjusting for age, gender and baseline CVRD status. Over a median of 6 years, current smoking, poor HbA1c control and hypertension conferred an additional 11.7%, 43.6% and 16.7% risk on MARCE. PAR estimates for the hypothetical modification of smoking, HbA1c or BP to reference categories are 4.0%, 3.6% and 2.8% for MARCE. Table 3 shows associations between risk factors and MARCE, and associated PAR.

Compared with those with more than 1 modifiable health risk behaviours, achieving ideal cardiovascular health could substantially reduce MARCE in individuals with T2D by an estimated 41.5%. In those with ≥3 and < 3 modifiable health risk behaviours, achieving ideal cardiovascular health could prevent 17.2% and 23.6% of MARCE in the T2D population, respectively. Other CVRD outcomes are in Additional files Table S1-S9.

Prevention of particular microvascular and macrovascular complications is already a priority in high-risk individuals with T2D [5, 19]. More generally, there are increasing calls for prioritisation of non-communicable diseases and multi-morbidity during and after the pandemic, whether to reduce risk of severe COVID-19 or the overall global burden of disease [11, 3]. However, the majority of clinical and public health efforts rest on short-term risk, and even in prevention, 10 years is the maximum window over which risk is estimated and conveyed to patients. However, focus on acute and shorter-term outcomes under-estimates total disease burden, where nearly all individuals with T2D and previous CVRD go on and have further MARCE over their lifetime. Improved secondary prevention is a major opportunity to reduce healthcare costs to patients and health systems. Given high lifetime T2D risk globally [20‐22], the high lifetime risk of CVRD which we describe, prevention earlier in the disease spectrum is of relevance to researchers, clinicians and policymakers, even in those with T2D alone (which is considered lower risk by existing prediction tools [23, 24]). For example, our findings may change the misperception of the risk status in the younger T2D population and inform change in national guidelines for diabetes and cardiorenal risk management in these individuals.

Compared with other diseases, those with T2D have a high lifetime risk of CKD (1 in 2), HF (2 out of 3) and CVRD (at least 8 out of 10). Research and practice in T2D and other long-term conditions emphasise multi-morbidity at baseline, but have neglected development of multi-morbidity over the life course [25], where benefits of primordial and primary prevention are greater for individuals, populations and health systems. The majority of scientific and clinical discourse has been disease-specific, with multi-morbidity focusing on number rather than specific combinations of diseases and risk factors. Studies have concentrated on coronary disease or CVD specifically [5]. Whether traditional statistical methods (e.g. regression analysis) or novel analytics such as clustering to discover subtypes or predict risk, studies have tended to consider individual diseases rather than groups of diseases together, such as CVRD [26]. Perhaps truly unsupervised approaches need to look at wider ranges of disease or composite exposures (e.g. CVRD) and outcomes (e.g. MARCE) over longer periods of time, if not over the life course, in order to more accurately reflect subtypes, distributions and trajectories of disease, with implications for chronic disease prevention and management.

Cardiorenal [26] and cardiometabolic syndromes [5] have gained attention in recent years from diagnosis to treatment. We show a particular risk of CKD, suggesting future prioritisation for earlier preventive therapies. Our data support recognition of CVRD, MARCE and broader outcomes for clinical practice, public health and research, from trials to service delivery. A recent long-term study of individuals with HF in the UK showed high prevalence, hospitalisation and mortality in presence of T2D and CKD [27]. Median survival was lowest in individuals with HF, CKD and T2D (0.7–2.8 years depending on stage of CKD), compared with HF-T2D-only (4.1 years), HF-CKD only (2.2 years) and no CKD or HF (4.4 years). We now add importance of burden of HF in T2D, partly due to high lifetime risk (30%), but also because 75% of MARCE are attributable to HF in those with T2D and HF. Research and clinical guidelines in T2D have stressed atherosclerotic CVD, but must also emphasise primary and secondary prevention to CVD more broadly, particularly HF. Health systems have responded by specialisation, when greater integration of specialties, more emphasis on primary care, early identification, recognition of multi-organ impairment and treatment of the people at most risk at multiple stages in the lifelong trajectory of T2D are required. Joint guidelines across societies and disease specialties are a positive step, but concerted collaboration is required, with relevance to other long-term conditions

Given common risk factors across common diseases, integrated primary prevention approaches such as “ideal cardiovascular health” and the American Heart Association’s “Life’s Simple 7” are likely to have a greater impact than disease-specific approaches. Ideal cardiovascular health in individuals with CKD is associated with a lower risk of end-stage renal disease and mortality [19]. Our analyses in T2D show the importance of these factors and behaviours in the prevention of CVRD, to inform public health messages to patients.

Novel therapies such as SGLT-2 inhibitors have shown improved outcomes across organ systems in T2D [9], renewing interest in overlap between HF, CKD and T2D. However, we confirm practice gaps in utilisation, despite established and evidence-based therapies for CVD prevention in people with T2D. For example, only two-thirds and half of the T2D patients with atherosclerotic CVD were taking statins and aspirin, respectively. Improved drug therapy, smoking cessation and blood pressure and glycaemic control are crucial targets at individual and population levels to reduce CVRD in the context of T2D.

Although previous studies have considered lifetime risk of CVD, CKD and CVD in T2D [14‐16, 28], this is the first study to date of lifetime risk of CVD and CKD together, and the first study across these diseases in T2D. We used lifetime risk, which gives a more complete picture of total disease burden, accounts for the effect of prolonged exposure to risk factors, remaining life expectancy and competing causes of death. Our analyses used nationally representative, population-based health record data, increasing the generalisability of our findings. There are some limitations. In the UK, our study population was 5% of the overall population. We did not have country-level data outside the UK and so our findings may need to be validated in other countries. Our analyses of risk use retrospective EHR data. In this study, we were unable to include asymptomatic CVRD or the incidence of CVRD without the need for hospitalisation in the outcome. The criteria for hospitalisation for CVRD may vary by clinician and institution and this was neither captured nor assessed in our analysis. We modelled using limited comorbidities, and absolute risk estimation methods do not allow for covariate adjustment throughout the lifecourse. For example, we did not study the impact of ethnicity. Residual confounding attributable to variables not covered by our data, such as diabetes duration, proteinuria and assessment of cardiac function might have influenced the results.