Individuals diagnosed with ASCVD from 1994 onwards were identified by ICD-10 codes or procedural codes in the Danish National Patient Register (NPR) (22). Inclusion criteria are presented in Table 1. Within the population of individuals with ASCVD, CKD stages 3–4 was determined by eGFR in the range 15–59 mL/min/1.73 m², as reported in the RLRR (23). Finally, individuals were required to be observed via at least two CRP tests fitting the criteria outline above, within six months, in the two years before or after index date.

The index date was defined as the date of diagnosis of ASCVD if the individual already had a diagnosis of CKD or the date of diagnosis of CKD if the individual already had a diagnosis of ASCVD.

The RLRR contains test results from 2011 onwards, with some regional variation in reporting. To ensure that only incident individuals were included in the study, all individuals diagnosed with CKD in the first year of reporting in the RLRR were excluded. Similarly, individuals diagnosed with ASCVD prior to, or in the first year of, reporting in the RLRR were excluded. Some hospital regions only started reporting after 2011. The first potential year of reporting for the individual was therefore decided by the region of residence, coinciding with when the region in question started reporting to the register. Consequently, the study population consisted exclusively of individuals newly diagnosed with both ASCVD and CKD from 2012 onwards.

Conditions of an inflammatory or a chronic nature are also thought to increase CRP levels. Hence, individuals with a diagnosis of HIV, hepatitis C, tuberculosis, inflammatory bowel disease or cancer, or with indications for transplant, were excluded. Criteria for identifying these conditions are provided in Supplementary Table S1.

CRP test results were obtained from the Register of Laboratory Results for Research (RLRR). In Denmark, CRP is frequently included in routine blood tests performed in both primary care and hospital settings, meaning that many CRP measurements are obtained as part of general clinical evaluation. As CRP test results were obtained only when individuals sought healthcare services, and due to the mode of testing, however, some CRP test results had to be excluded.

It was not possible to distinguish high-sensitivity CRP tests from wide-spectrum CRP tests. However, modern laboratory equipment is capable of accurately identifying CRP levels <10 mg/L, despite using wide-spectrum CRP (wsCRP) assays (20). Two common CRP test methods (assays) used during the study period had a lower range of CRP = 2.9 mg/L and CRP = 4 mg/L. All tests corresponding to these values were therefore excluded, as they likely did not represent the true CRP level.

Furthermore, test results could have been observed in association with infections requiring antibiotic or antiviral medications. Hence, all tests observed seven days before or 30 days after collection of a prescription for antibiotic or antiviral medications were excluded. Data on the collection of prescriptions were obtained from the Register of Medicinal Product Statistics (21).

Moreover, ASCVD events can trigger an immune response which results in increased CRP. Hence, all test results observed in the 30 days following an ASCVD event were excluded.

SI has been defined as having one high-sensitivity CRP test in the range of 2 mg/L–20 mg/L (24, 25). This definition is applicable in screening studies or inclusion criteria for a randomised controlled trial. In this study, however, individuals are tested when seeking medical care. As there are multiple reasons CRP might be elevated when seeking medical care, an alternative criterion to define SI had to be applied.

Hence, to be as certain as possible that individuals included in this study actually have SI, SI was defined as individuals having at least two CRP tests in the 2 mg/L–20 mg/L range less than six months apart. The same requirement of having multiple CRP tests were applied for the group without SI as well, to mitigate the issue of potential misclassification due to transient elevations in CRP and to mitigate the potential issue of differences in testing frequency.

Furthermore, since the exact incidence date of SI cannot be determined, the six-month time period had to be observed within two years of the study index date.

Dates of death occurring between 1 January 2011 and 31 December 2022 were gathered from the Danish Register of Causes of Death (26). As a secondary outcome, the time from index date to either first ‘MACE or death’ was applied. MACE was defined by ICD-10 codes in the NPR as myocardial infarction (ICD-10: I21, I22, I23, including subgroups) or ischemic, haemorrhagic or undetermined stroke (ICD-10: H34.1, G45–46, I60–61, I63, I64) as primary diagnosis, or as all-cause mortality following the latter incidence of ASCVD or CKD (27, 28). As all individuals had established ASCVD at baseline, MACE events reflect subsequent (i.e., recurrent) cardiovascular outcomes.

Moreover, the results are put in context by an investigation of the presence of selected comorbidities at index date, and of the incidence of select comorbidities following index date. The select comorbidities were stroke, heart failure, type 2 diabetes, type 1 diabetes, chronic obstructive pulmonary disease, myocardial infarction and atrial fibrillation. The method of identification for each comorbidity is presented in Supplementary Table S4.

The Cox proportional hazards model was used to quantify the hazards associated with SI following index date. The exposure of interest was SI (yes/no). The effect was estimated for the population as well as for subgroups. Hazard ratios for subgroups were estimated through the interaction of SI with the relevant subgroup indicator and by calculating the effect in the group using delta methods (29).

The Cox model was adjusted to avoid potential confounding in the SI/no-SI groups. The included covariates were separated into demographic, biomarker and comorbidity categories. A forward-stepping approach was applied to observe whether any of the categories had a major impact on the hazard related to SI. The final model was adjusted for age, sex, region of residence, education, history of heart failure, diabetes, triglyceride, HbA1c and cholesterol (LDL-c). Sensitivity testing, including additional adjusting for cardiovascular diseases (CVD) such as hypertension and atrial fibrillation, did not influence the inference from the model.

The Cox model validity was assessed by Schoenfeld Residuals. To ensure that the proportional hazard assumption was upheld, individuals were required to survive at least 10 days after index date to be included in the study.

The same approach was applied for the ‘MACE or death’ outcome. Cumulative probability plots for ‘MACE or death’, respectively, are presented alongside the hazard ratios.

Because the study contains up to a 10-year follow-up time, an individual's SI status may change from index date to the end of the follow-up period. To ensure that the results are not a consequence of SI classification timing, a model using repeated measurements of CRP is also presented. The model specification reflects the static model, but model residuals were clustered for each individual. In this model, the CRP values were stratified by categories <2 mg/L (reference), 2–<5 mg/L, 5–<10 mg/L and 10–<20 mg/L.

Finally, all analyses were conducted using two alternative definitions of SI: (1) a relaxed criterion, classified as one CRP test result of 2 mg/L–20 mg/L within two years of index date as SI, and (2) strict criteria, classified as three CRP test results of 2 mg/L–20 mg/L less than six months apart, within two years of index date, as SI.

Of the 19,159 patients, 13,036 (68%) had SI according to the definition of two valid measurements with values between 2 mg/L and 20 mg/L. In the no-SI population (n = 6,123, 32%), 30% of the CRP test results were above 2 mg/L.

The SI and no-SI populations did not have clinically significant differences with respect to age, sex, region of residence, education or cultural background (Table 2). Both groups included slightly more men than women, and most individuals were diagnosed with ASCVD prior to being diagnosed with CKD. The SI population had a higher burden of illness at index date, with more frequent history of hypertension, atrial fibrillation, COPD, heart failure and type 2 diabetes. This population did not have a significantly different history of stroke or acute myocardial infarction.

Table 3 presents the hazard from the Cox regressions. The table presents the results from an unadjusted model as well as the final model for the overall effect and by subgroups. The table contains the HR for mortality as the main outcome, and the HR for ‘MACE or death’. Hazard ratios from stepwise models are presented in Supplementary Table S2. Kaplan–Meier plots and cumulative probability distributions for MACE and all-cause mortality are presented in Figures 2a,b. The cumulative probability distributions are based on a competing risk approach, allowing separate visualisation and interpretation of MACE and mortality.

In the final multivariate-adjusted model, SI was associated with a 106% higher risk of death (HR: 2.06, CI: 1.92–2.21) and the hazard of ‘MACE or death’ by 66% (HR: 1.66, CI: 1.57–1.77). SI was associated with a higher risk of death and MACE in all analyses and subgroups; however, a significant difference in the risk of SI associated with mortality and MACE was estimated within the subgroups.

The results from the repeated measurement of Cox regression to account for potentially changing inflammation status during the study are presented in Figure 3. There is a clear gradient demonstrating increasing hazard of all-cause mortality as CRP levels increase. Compared to CRP <2 mg/dL, CRP of ≥2 mg/L to <5 mg/L, CRP of of ≥5 mg/L to <10 mg/L and CRP ≥10 mg/L increased the hazard of death by 70% (HR: 1.7, CI: 1.52–1.90), 140% (HR: 2.4, CI: 2.17–2.66) and 356% (HR: 4.56, CI: 4.14–5.02), respectively.

In Supplementary Tables S3, S4 the results using alternative definitions of SI is presented. In Supplementary Table S1, SI is defined by having one CRP test result of 2 mg/L–20 mg/L within two years of index date. In Supplementary Table S2, the analysis was based on an SI definition where three CRP test result of 2 mg/L–20 mg/L less than six months apart was observed within two years of index date. The results from these analyses does not change the inference we make from the results of the main analysis.

A total of 104,532 valid CRP test results were observed during the study period. Individuals with SI were tested about three times more frequently than were those with no SI. To investigate whether this difference in test frequency influenced the results, a model was developed that included the variable ‘number of tests.’ The effect was statistically significant, but with a hazard ratio (HR) of 1.03 [CI: 1.03–1.03] which is low compared to the found hazard ratios in the study.