The Hong Kong CAD study is a single-center prospective observational study of patients with stable CAD and was approved by the local Institutional Review Board [6‐8]. Patients who were > 18 years old and had stable coronary artery disease newly diagnosed by conventional angiography at Queen Mary Hospital between 1st December 2003 and 30th November 2014 were recruited. Those with impaired left ventricular ejection fraction < 50%, moderate or severe valvular heart disease, acute coronary syndrome or who were unable to or refused to provide informed consent were excluded. Clinical information on admission including age, gender, past medical conditions, family history, cardiac biomarker levels, electrocardiographic findings, echocardiographic data, medications prescribed, coronary angiography reports and revascularization results were recorded at baseline. All patients were prescribed statin prior to admission for coronary angiography and continued indefinitely after the diagnosis of CAD was confirmed. Data on their subsequent clinical course including medication usage, dosage and adverse effects were retrieved from the comprehensive electronic medical system of the Hong Kong Hospital Authority as previously described [17‐19]. Patients with incomplete clinical records, insufficient plasma sample for analysis, or those who refused or were not prescribed a statin on discharge were excluded.

In this study, SI was defined as an inability to continue use of a statin, either because of side effects, or because of elevated liver enzymes or creatine kinase that were sufficiently abnormal to cause concern [7]. Cases of SI were retrospectively ascertained using a standardized protocol (Additional file 1). In brief, patients were considered completely SI if they were unable to tolerate the initial dose of one statin and any dose of another statin [7]. Patients were considered partially SI if they were unable to tolerate at least one statin at one dose [7].

All subjects had plasma stored at − 80 °C at baseline. Serum level of hs-cTnI was retrospectively determined for this study after completion of patient recruitment, using a Chemoluminescent Microparticule ImmunoAssay (Architect i1000SR Abbott®, Paris, France) according to the manufacturer’s protocol. The level of detection was 1.2 ng/L and the 99th percentile value of serum hs-cTnI level in male and female control subjects was 8.5 ng/L and 7.6 ng/L, respectively [17‐19]. Based on a baseline sample in 20 patients, the intra-assay coefficient of variations was 2.5%.

Patients were monitored until the diagnosis of a cardiovascular endpoint, death or last clinical visit. The primary endpoint of this study was new-onset major adverse cardiovascular event (MACE). We adjudicated new-onset MACE based on the International Classification of Diseases, Ninth Revision (ICD-9): acute myocardial infarction (ICD-9 410), acute coronary syndrome (ICD-9 411.1), stroke (ICD-9 430, 431, 433, 434, 436), peripheral vascular disease (ICD-9 443.9) and cardiovascular death (death certificate ICD-9 410–447). Information on discharge diagnosis and date of events were confirmed by reviewing electronic records of the Hong Kong Hospital Authority by two independent cardiologists [17‐19]. For patients who died during follow-up, the main cause and date of death were obtained from the Hong Kong Death Registry.

Results are expressed as mean ± standard deviation or number and percentage as appropriate. Normality assumption was evaluated using the Kolmogorov-Smirnov test. The biomarker hs-cTnI was logarithmically transformed before analysis. Patient age was categorized as < 65 or ≥ 65 years. The optimal cut-off value for hs-cTnI was determined using the Youden J index. Cox proportional hazards regression was performed to examine the association of age, clinical risk factors, statin intolerance and hs-cTnI with MACE. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated for MACE. Additional variables including advanced age, male sex, active smoking, hypertension and diabetes were selected using minimized Akaike information criterion to build the multivariate model [20]. Kaplan-Meier survival curve and log-rank test were used to assess the relationship between SI, hs-cTnI and MACE. A two-sided P-value < 0.05 was considered statistically significant. All analyses were performed using statistical software packages SPSS (version 19; SPSS, Chicago, IL), STATA (version 14.0) and R-programming language (version 3.5.2).

Our cohort consisted of 1202 patients who had stable CAD diagnosed by conventional coronary angiography. Patients were excluded if they had missing clinical data (n = 33) or biomarker measurement (n = 214), or refused or were not prescribed a statin on discharge (n = 3). A total of 952 patients were retained in the analysis and their clinical demographic features are summarized in Table 1. We identified 138 (14.5%) patients who suffered either complete SI [13 (1.4%)] or partial SI [125 (13.1%)]. Two patients concomitantly received ezetimibe at baseline. Among those with SI, only two (1.5%) were due to significant elevation in creatine kinase, one (0.7%) was due to significant elevation in liver enzymes, and the remainder were due to symptoms attributed to statin use. Patients with complete SI were treated with ezetimibe (n = 1), fibrate (n = 3), or no lipid lowering therapy (n = 9). Patients with partial SI were treated with a reduced dose of statin (n = 7) or an alternative statin (n = 100), and add-on ezetrimibe (n = 7) or fibrate (n = 11).

Of note, patients with SI were more likely to have diabetes mellitus [61 (44.2%) vs. 275 (33.8%), P = 0.02] and have a higher hs-cTnI (2.6 ± 1.6 vs. 2.1 ± 1.3 ng/L, P < 0.01) at baseline than those without SI. Nevertheless there were no significant differences in LDL-C level, proportion of patients with LDL-C < 1.8 mmol/L, type or dose of statin between patients with or without SI (all P > 0.05. Table 1). At follow-up, although the total cholesterol (3.9 ± 1.0 vs. 3.7 ± 0.7 mmol/L, P < 0.01) and mean LDL-C level (2.0 ± 0.8 vs.1.9 ± 0.6 mmol/L, P = 0.01. Table 1) was significantly higher in patients with than those without SI, there were no significant differences in the proportion of patients with LDL-C < 1.8 mmol/L, the type of dose of statin between patients with and without SI (all P > 0.05, Table 1).

After a mean follow-up of 51 ± 42 months, 148 (15.5%) patients developed a MACE, including 54 (5.7%) acute coronary syndrome; 19 (2.0%) stroke; 9 (0.9%) peripheral vascular events and 66 (6.9%) cardiovascular deaths. Clinical characteristics of CAD patients with and without MACE are summarized in Table 2. Stable CAD patients with MACE on follow-up were older, and a higher proportion were male or had hypertension or diabetes compared with those without MACE (Table 3, all P < 0.05). Moreover, patients who developed a MACE were more likely to have SI [41 (27.7%) vs. 97 (12.1%), P < 0.01] than those who did not (Table 2). Kaplan-Meier MACE-free survival showed a significant difference between stable CAD patients with and without SI (log-rank P < 0.01. Figure 1a).

As shown Table 1, hs-cTnI level was significantly higher in CAD patients with SI than those without (2.6 ± 1.6 vs. 2.1 ± 1.3 ng/l, P < 0.01). Moreover, hs-cTnI level was significantly higher in CAD patients with MACE than those without (3.1 ± 1.6 vs. 2.1 ± 1.2 ng/l, P < 0.01). The cut-off value of hs-cTnI for MACE derived from the Younden J index was 5.2 ng/l. Multivariable Cox regression analysis showed that both SI (HR 1.52, 95% CI 1.06–2.19, P = 0.02) and elevated hs-cTnI level (HR 3.18, 95% CI 2.07–4.89, P < 0.01) were independent predictors of MACE in patients with stable CAD (Table 3). Nevertheless, when stratified by hs-cTnI level, SI independently predicted worse MACE-free survival in patients with elevated hs-cTnI (HR 1.51, 95% CI 1.01–2.24, P = 0.04), but not in patients without (HR 1.40, 95% CI 0.55–3.52, P = 0.48. Table 4). The MACE-free survival was significantly worse in stable CAD patients with both SI and elevated baseline hs-cTnI than those without SI or elevated hs-cTnI (log-rank P < 0.01. Figure 1b).

First, this was a single-center cohort study that included only CAD patients treated with statin, which inadvertently resulted in selection bias over patients who could tolerate the initiation of statin therapy. As a result, we might have underestimated the prevalence of SI in our population. Second, some patients in this study were recruited over 15 years ago, when the benefit of high-dose statin and low LDL-C in stable CAD was less well-established. Since stable patients are often continued with the same statin dose after initial prescription [9, 10], a significant proportion of patients in this study remained on low dose statin. In fact, only 43% of patients without SI in this study achieved a target LDL-C level of < 1.8 mmol/L on follow-up, which was similar to those observed in other Asian countries including Taiwan and Japan [9, 10]. As the latest guidelines recommended titration of both statin and non-statin therapies to achieve a 30–49% reduction in LDL-C [5], it remains unclear whether the adoption of this approach will alter the results of this study. Third, retrospective assessment of hs-cTnI in stored serum samples using immunoassays may result in random measurement errors. It is because the long-term stability of cTnI in frozen sample has remained debatable [27‐30], together with the presence of heterophile or anti-TnI antibodies, may lead to erroneous hs-cTnI values [31]. Nevertheless, in a recent study that assessed cTnI levels using the Architect STAT hs-cTnI assay (Abbott Laboratories, IL, USA) in serum samples that were stored for 0.8 to 82.0 months, the mean cTnI value in the quartile of samples with the longest storage time did not differ significantly from those with the shortest storage time, suggesting that the magnitude of cTnI degradation over a long period of storage time may be small compared with the actual cTnI value [32]. Moreover, how the presence or absence of antibodies interfere with the prognostic significance of hs-cTnI values remain unknown [31]. Despite these limitations, hs-cTnI has been shown to predict clinical outcomes in a variety of cardiovascular conditions [13‐15], which is consistent with the findings of this study. Before the development of a simple and low-cost test that is negatively influenced by those antibodies, our findings is still of significant value as hs-cTnI measured by immunoassays is widely available in the clinical setting.