SWEDEHEART is a Swedish national register in which patients with acute coronary syndrome are prospectively registered. Patient characteristic, hospital treatments, drug treatments at discharge, and outcome for patients consecutively included and treated at all Swedish coronary care units are collected in this register [13]. The Swedish Secondary Prevention after Heart intensive care Admission (SEPHIA) is a sub register within SWEDEHEART collecting data on secondary prevention and cardiac rehabilitation [14]. Follow-up data were registered by office visits or phone call, supplemented by blood samples collected at the patient’s primary care centre, at six to ten weeks and at 12 to 14 months post MI. Around 80% of all Swedish MI patients below the age of 75 years are included in this register [14].

In this study, a cohort of 5904 patients (74% men) registered in the SEPHIA register and who had one year follow-up during 2013, were included. Data from the SWEDEHEART/SEPHIA national register was extracted in an aggregated form. In accordance with Swedish regulations, written informed consent is not necessary for national registers, however all patients were informed about their participation in the register, and their right to decline participation.

The Reduction of Atherothrombosis for Continued Health (REACH) risk function was used to predict 20-month risk of recurrent CVD [15]. This model provided estimates of recurrent non-fatal and fatal CVD events based on the following risk factors: age (years), gender (male/female), smoking, diabetes mellitus; body mass index (BMI) <20 kg/m², number of vascular beds with CVD-manifestations (1, 2, 3), congestive heart failure, atrial fibrillation, statin treatment, and acetylsalicylic acid treatment [15]. The predicted 20-month fatal and non-fatal CVD risk were derived from the “next event” REACH equation using the detailed Cox regression model covariate coefficient estimates provided in Wilson et al. 2012 Appendix [15]. CVD rates were predicted separately for males and females, before calculating a weighted average the overall cohort risk, accounting for gender variation in risk factors. CVD risk was estimated separately for each year in the prediction time period, accounting for the yearly increase in cohort age and the impact of increased age on CVD risk. The predicted 20-months risks were annualized for each year of prediction. The effect on CVD risk derived from the lowering of LDL-C was calculated based on the Cholesterol Treatment Trialists’ Collaboration (CTTC) meta-analyses results, linking LDL-C lowering to CVD event risk reduction [6]. Different rate ratios of CVD event reductions per mmol/L LDL-C reduction were used based on CTTC: MI (0.71), ischemic stroke (IS) (0.69) and fatal CHD (0.80) [6]. For fatal stroke a rate ratio of one was used based on the non-significant difference reported by CTTC [6]. The proportion (%) of non-fatal (MI and stroke) vs. fatal CVD event post MI, was based on Jernberg et al. reported event distribution of up to 24 months post MI: 46.8% CVD death, 37.8% MI, 15.4% stroke [16]. In the age ranges 55–64 to 65–74 years of fatal CHD vs. fatal stroke occurred in 96% vs. 4% post-MI, indicating that fatal CHD is more common than fatal strokes in post MI patients [17]. The direct costs of non-fatal MI and stroke were based on Hallberg et al. 2015 [18] and were for MI: Swedish crowns (SEK) 76,657, and ischemic stroke: SEK 88,790. These event cost estimates were from Table 4 in Hallberg et al. 2016, and from the incremental cost year (day 0–365 days after new CVD event) for the CVD history cohort. The reported cost estimates were converted to SEK using the same exchange rate of 1 € = 8.71 SEK as reported by Hallberg et al. 2015, p. 3. Fatal CVD costs were estimated, and based on Ara et al. 2009 [19] and were for CHD death costs SEK 11,345 (14.8% of MI costs) and stroke death SEK 40,577 (45.7% of stroke costs). Total directs costs of CV events were estimated in a first analysis step. In a second step the total cost including costs of informal care by family and relatives, indirect costs of productivity loss due to premature death, and reduced work capacity ere estimated based on previous findings have shown that direct costs accounted for around 41% of total costs [20].

The REACH risk prediction model was based on participants from around the world with different prior CVD events, not only MI [19]. Results from the UK Clinical Practice Research Datalink (CPRD) calibration analyses indicated that the REACH risk prediction significantly underestimated the risk of CVD events in a post-acute coronary syndrome population [21]. Analyses of the REACH risk prediction were therefore, calibrated according to CPRD analyses. The CPRD study included heart failure (HF) in addition to MI, stroke and CVD mortality outcomes, and hence the reported calibration factor of 3.36 for a post- acute coronary syndrome (ACS) population had to be adjusted for the purpose of this study. Based on the post-ACS cohort, the adjusted calibration factor was 3.06 = 3.36*(1–0.089) in the CPRD cohort accounting for HF incidence in patients between 64 and 73 years old. In addition to the prediction of fatal and non-fatal CVD event using the REACH risk prediction, the predictions account for Swedish age- and gender specific non-CVD mortality life tables from Statistics Sweden were used (available at www.​scb.​se/​hitta-statistik).

Demographics and other baseline characteristics were presented for the overall study population, as well as for the controlled cohort (LDL-C ≤ 1.8 mmol/L), the non-controlled (LDL-C > 1.8 mmol/L) cohort, and in men and women separately. The REACH risk function was used to predict the CVD risk [15], as well as the calibrated CVD risk prediction described above [19]. The possible avoided costs were based on cases at baseline in REACH (henceforth called base case), the calibrated (scenario) risk predictions, the corresponding population’s CVD costs and potential cost reductions linked to LDL-C reduction according to guidelines. This was predicted by combining event prediction and estimation of health care costs associated with each type of event.

The presented analyses and results focused on the assessment of the impact on CVD risk and costs if the current LDL-C target was reached in comparison with the current treatment patterns and practices. CVD risk and costs were predicted over a 10-year time horizon accounting for increased age and consequent increased CVD risk over time. However, the analysis was not accounting for any other risk factor changes over time, and is furthermore not accounting for possible future change of statin treatment goal and attainment rate which may change further with physicians’ care or disease progression. The post-MI population in this study was younger, due to the age limit of the SEPHIA-register, more often revascularised, and had a higher proportion of men, also due to age inclusion criteria, compared with the MI population in general, which limited the generalizability. Since ICD-10 codes were used for morbidity data there was a possibility for coding errors. However, a validation of the Swedish in patient register showed that coding was concurrent in >98% of the cases [32]. The study was based on population averages, and not individual patient-level data. For the analysis of implementation of the guidelines focused on LDL-C this poses some challenges as the guidelines specifies that target LDL-C is ≤1.8 mmol/L, or if this cannot be achieved, a 50% reduction in LDL-C levels. Recently European Guidelines on CVD in clinical practice recommended an LDL-C goal <1.8 mmol/L, or a reduction of at least 50% if the baseline is between 1.8 and 3.5 mmol/L in very high risk patients [33]. The patients in this study represent around 80% of the total MI-population in Sweden in this age group and, potential selection bias was considered to be low. There is a risk of conservative bias, as patients with lower social, financial and health-related functioning may be less likely to attend follow-up visits. The study population should however provide a good estimate of treatment strategies and treatment goals attainment in usual care. Possible future research should compare these risk model based on predictions with actual real-world outcomes over a longer period in the SEPHIA register.