Inflammatory response by 48 h after admission and mortality in patients with acute myocardial infarction complicated by cardiogenic shock

Abstract

Aims

Cardiogenic shock (CS) is known to induce an inflammatory response. The prognostic utility of this remains unclear. To investigate the association between C-reactive protein (CRP) levels and leucocyte count and mortality in patients with acute myocardial infarction complicated by CS (AMICS).

Methods and results

Consecutive patients (N = 1716) admitted between 2010 and 2017 with an individually validated diagnosis of AMICS were included. The analysis was restricted to patients alive at 48 h after first medical contact and a valid CRP and leucocyte measurement at 48 ± 12 h from the first medical contact. A combined inflammatory score for each patient was computed by summing the CRP and leucocyte count z-scores to normalize the response on a standard deviation scale. Associations with mortality were analysed using a multivariable Cox proportional hazards model stratified by inflammatory response quartiles: Of the 1716 patients in the cohort, 1111 (64.7%) fulfilled inclusion criteria. The median CRP level at 48 h was 145 mg/dL [interquartile range (IQR) 96–211]. The median leucocyte count was 12.6 × 10⁻⁹/L (IQR 10.1–16.4). Patients with the highest inflammatory response (Q4) had lower median left ventricular ejection fractions and higher lactate levels at the time of diagnosis. The 30-day all-cause mortality rates were 46% in Q4 and 21% in Q1 (P < 0.001). In multivariable models, the inflammatory response remained associated with mortality [hazard ratio (HR)_Q4 2.32, 95% confidence interval (CI) 1.59–3.39, P < 0.001]. The finding was also significant in AMICS patients presenting with out-of-hospital cardiac arrest following multivariable adjustment (HR_Q4 3.37, 95% CI 2.02–4.64, P < 0.001).

Conclusion

Cardiogenic shock induces an acute inflammatory response, the severity of which is associated with mortality.

Graphical abstract

Open in new tabDownload slide

Introduction

Cardiogenic shock (CS) is a critical condition that leads to a variety of detrimental effects on end organs following circulatory failure. It has been reported in 5–10% of cases of acute myocardial infarction (AMICS) and has a high short-term mortality of 40–60%.^1–3 Despite changing characteristics with regard to the use of percutaneous coronary intervention, mechanical circulatory assist devices, and mechanical ventilation, mortality has improved minimally during the last two decades.^4,5

The pathophysiology underlying AMICS is complex; however, reduced cardiac output following the initial myocardial injury may initiate a vicious cycle of haemodynamic deterioration and ultimately circulatory collapse if not treated in time. This is reflected in the definitions of CS, which are based on signs of hypoperfusion according to various degrees of end-organ dysfunction or damage, such as in Society for Cardiovascular Angiography and Interventions (SCAI) grading.⁶

The systemic nature of CS induces an inflammatory response, which may contribute to the pathophysiology of further circulatory decline.⁷ Higher levels of circulating markers of systemic inflammation such as C-reactive protein (CRP) have long been associated with a higher risk of myocardial infarction and stroke, and elevated leucocyte count on admission in patients with myocardial infarction has been associated with higher short-term mortality.^8,9 Evidence of an exaggerated inflammatory response has been associated with poor prognoses in out-of-hospital cardiac arrest (OHCA), heart failure, and CS patients previously.^10–13

Schemes that integrate clinical observations with blood biomarkers exist with the purpose of detecting early sepsis, which itself has a principal inflammatory component and is particularly dangerous in nature with complications of shock and mortality rates upwards of 35–40%.¹⁴ Definitions such as Systemic Inflammatory Response Syndrome (SIRS) are based on body temperature, heart rate (HR), respiratory rate, and leucocyte count, while Sequential Organ Failure Assessment Score adds markers of liver and renal dysfunction.

Procalcitonin (PCT) is an emerging marker for further differentiation of such responses.¹⁵ The occurrence of infection secondary to CS is plausible considering the general prevalence in critical care settings; however, diagnosis is challenging when indicator biomarkers are elevated at baseline.¹⁶

The aim of the present study was to investigate the systemic inflammatory aspect of AMICS and determine whether elevated levels of routine inflammatory biomarkers during admission were associated with 30-day mortality in patients with AMICS.

Methods

Study population

The study is based on the RETROSHOCK cohort which comprised all patients with AMICS from 1 January 2010 to 31 December 2017, admitted at any hour to any of two tertiary Danish heart centres providing AMICS treatment for two-thirds of the Danish population (3.9 million citizens). Retrospective screening was performed using national Danish health registry records of admissions, diagnoses, and procedures. The screening algorithm was based upon the primary/secondary ICD-10 diagnosis code: CS (R57.0) and/or AMI (I21.x) and/or cardiac arrest (146.x), the latter two with additional criteria of death during admission, intensive care unit (ICU) treatment, vasoactive drug treatment and/or mechanical assist device treatment. Discharge summaries of screened patients were individually reviewed by the authors to verify AMI diagnosis in accordance with the universal definition of myocardial infarction.^17,18 A full chart review was performed for verified AMI patients, applying CS criteria, thus defining a final AMICS population (N = 1716). The cohort was compiled using baseline demographics and clinical parameters recorded on admission, including angiographic data and the recorded use of mechanical ventilatory and circulatory support. Follow-up data on 30 days all-cause mortality were available for all patients. A detailed description of cohort establishment has been previously published.²

Definition of acute myocardial infarction complicated by cardiogenic shock

Cardiogenic shock was determined in screened AMI patients if all the following was present: (i) persistent hypotension (systolic BP ≤ 90 mmHg for >30 min and/or use of vasoactive medication), (ii) signs of organ hypoperfusion (altered mental status, cold skin, oliguria, lactate ≥2.5 mmol/L), and (iii) reduced right or left ventricular function in the absence of hypovolemia, sepsis, anaphylaxis, pulmonary embolism, or primary valve dysfunction.

Laboratory data

Clinical blood samples obtained during admission were analysed by each centre’s Department of Clinical Biochemistry, both of which had accreditation status as medical laboratories and reported the results electronically to electronic health record systems. The type and origin of time-stamped sample records were coded according to the national terminology that standardizes the systems exchange of laboratory records. Records matching national patient identification numbers within the period of admission were retrieved for CRP and leucocyte counts. Records that were not numerically interpretable were discarded, as in the case of sample handling or analysis errors. Each CRP and leucocyte value was coded as the duration in minutes from the patient’s first medical contact. If more than one record was present within the measurement window, the highest numerical value was used. Data for high-sensitivity cardiac troponin (hs-cTn) and PCT were retrieved in a similar manner, where available.

Inflammatory score definition

For biomarker classification, only patients alive beyond admission day number two were subject to analysis, who also had at least one CRP and leucocyte count measured at 48 ± 12 h from the first medical contact. The 2-day landmark was chosen to allow for an inflammatory response to establish and the previously described initial days’ high mortality rate to stabilize.¹⁹

C-reactive protein and leucocyte count values from patients who were subjected to analysis were individually standardized by computing z-scores that expressed sample deviation from the distribution mean in units of standard deviations (SDs), with z = (x − μ)/σ, where x = raw score, μ = distribution mean, and σ = distribution SD. The risk of immortal time bias was addressed by limiting all modelling and analysis to patients alive beyond the fixed 48 h time point; no clinical parameters were used prior to this step of analysis.

A combined measure of systemic inflammation for each patient was computed as the arithmetic sum of CRP and leucocyte z-scores. Combined scores were categorized into quartiles and used to compare baseline characteristics, major risk factors associated with AMICS, and outcomes. Analyses were repeated after log-2 transformation of the biomarkers, which did not significantly alter the conclusions. Another model evaluated PCT separately and as incorporated into the inflammatory score.

Inflammatory marker dynamics

For further characterization of inflammatory marker dynamics on each side of the primary analyses’ 48 h mark, the change was calculated for each patient by assessing levels at two points in time (T1, T2, ± 12 h), as has been described in adjacent settings. The exact time difference between samples Δt was used to express change in % as follows: (T1 − T2)/(T1 · Δt) × 100. The CRP change was analysed for Days 0–3, corresponding to the observed primary response pattern. Leucocyte change was analysed for Days 3–7 to characterize any secondary response occurring after the primary analysis. When multiple samples existed within each time window, the earliest and latest were used, respectively.

Analyses were restricted to survivors within each timespan to address survival time bias.

Statistical analysis

Continuous variables are reported as mean (standard deviation) when normally distributed and median (1st quartile–4th quartile) for non-normally distributed measures. Categorical variables are reported as frequencies (percentages). Pairwise comparisons adjusted for multiple testing were performed using the Tukey and Benjamini and Hochberg methods on normal and non-normal variables, respectively, with a reported P-value for trend.

Results are expressed as hazard ratios (HR) and 95% confidence intervals (CIs), and group mortality differences with a log-rank P-value. For all analyses, exact P-values above 0.001 were reported, with values <0.05 considered statistically significant.

Primary analyses were repeated for each centre (N = 434/N = 677) to indicate whether results were stable across the two differing analysing laboratories as well as local sampling and treatment practices. Models were determined to be stable across centres, resulting in similar hazard ratios and significance levels.

R version 4.2.2 (31 October 2022) was used for analysis and data management.

Primary outcome analysis

The combined z-score quartiles were used to estimate 30-day mortality using Kaplan–Meier analysis. The log-rank test was used to compare the survival distributions.

Associations with mortality were further analysed using univariable and multivariable Cox proportional hazards regression models. The latter were adjusted for age, sex, body mass index (BMI), arrival hs-cTn (z-scored), heart rate, systolic blood pressure (sBP), and left ventricular ejection fraction (LVEF) at CS diagnosis, mechanical ventilation, bacteraemia, and previous diabetes mellitus (DM). The selection of covariates was performed at the authors’ discretion, considering factors that are known or could be assumed to interact with mortality in the population and disease context. Variable inflation factors were assessed among covariates to identify multicollinearity before admission, addressing the risk of effect over-estimation.

Results

During the study period, 1716 patients were included in the cohort, with an overall 30-day mortality of 53.7%. A total of 1212 (71%) patients survived on Day 2 of admission. Of this total, 1130 (93%) had at least one CRP sample, while 1121 (92%) had at least one leucocyte count sample available within 48 ± 12 h from the first medical contact. The intersecting group of 1111 patients (92% of eligible patients) was classified (Figure 1). The characteristics of the patients not eligible for the primary analysis are available (see Supplementary material online, Table S1).

Individual biomarkers

The median CRP level among all patients was observed to peak on Days 2 and 3 (48–72 h) and leucocyte count on Days 0 and 1 (0–24 h) (Figure 2). The median CRP level at 48 h was 145 mg/dL [interquartile range (IQR) 96–211]. The median leucocyte count at 48 h was 12.6 × 10⁻⁹/L (IQR 10.1–16.4). The trend on the day of the peak was similar regardless of survival status on Day 2.

The 30-day mortality was significantly higher in the upper CRP level quartile than in the lower (47 vs. 24%, P_logrank < 0.001), as well as in the upper vs. lower leucocyte count quartile (45 vs. 24%, P_logrank < 0.001).

The upper CRP level quartile remained associated with mortality in the univariable (HR_Q4 2.23, 95% CI 1.66–3.00, P < 0.001) and multivariable models (HR_Q4 1.74, 95% CI 1.22–2.49, P = 0.002), both with reference to the lower quartile. Increased leucocyte count was also significantly associated with mortality in univariable (HR_Q4 2.25, 95% CI 1.68–3.02, P < 0.001) and in multivariable models (HR_Q4 1.98, 95% CI 1.40–2.81, P < 0.001).

Four hundred and twenty-three patients in the study cohort who had PCT assayed at 48 ± 12 h were identified; the upper response quartile was similarly associated with mortality (see Supplementary material online, Tables S2 and S3).

The median change in CRP between Days 0 and 3 was +76% (IQR 19–217). The highest increase (4th vs. 1st quartile) was associated with a lower mortality (HR_Q4 0.70, 95% CI 0.51–0.95, P = 0.025). The median change in leucocyte count between Days 3 and 7 was 0.13% (IQR −0.4 to 0.2). The highest increase (4th vs. 1st quartile) was associated with a higher mortality (HR_Q4 1.74, 95% CI 1.10–2.76, P = 0.017) (See supplementary material online, Figures S1–S3).

Combined inflammatory response

Baseline characteristics of the combined inflammatory response quartiles are presented in Table 1. In comparison to the lower quartile, the upper quartile was characterized by higher BMI, higher heart rate, and a trend towards higher lactate. Left ventricular ejection fraction was lower in the higher inflammatory quartiles, whereas extracorporeal membrane oxygenation (ECMO) was applied less. The SCAI grading was also more severe in the upper quartiles (P < 0.001, Table 1).

A higher combined inflammatory response at 48 h was associated with a significantly higher overall 30-day mortality and a longer median duration of ICU admission (Table 1). The response remained significantly associated with mortality in both univariable and multivariable models (Table 2, Figure 3). Adding admission lactate to the models did not significantly alter results.

When incorporating PCT into the combined response (N = 423), the association with mortality was attenuated in univariable (HR_Q4 1.90, 95% CI 1.22–2.79, P = 0.005) and multivariable analyses (HR_Q4 2.02, 95% CI 1.17–3.48, P = 0.01). Analyses were also performed for PCT separately (see Supplementary material online, Tables S2 and S3).

Out-of-hospital cardiac arrest

Out-of-hospital cardiac arrest prevalence did not differ across inflammatory quartiles. Out-of-hospital cardiac arrest patients (n = 565) had higher CRP at 48 h than non-OHCA patients [150 mg/dL (IQR 108–208) vs. 135 mg/dL (IQR 79–214), P < 0.001] and lower leucocyte count [11.9 × 10⁻⁹/L (IQR 9.4–15.4) vs. 13.7 × 10⁻⁹/L (IQR 10.9–17.7), P < 0.001]. In both ± OHCA subgroups, combined inflammatory status remained associated with mortality in the univariable and multivariable models (Table 3).

Discussion

In the present study, we found that higher levels of systemic inflammation at 48 h were associated with a significantly higher 30-day mortality. This was based on a large multicentre cohort of patients with individually validated AMICS diagnoses, coupled with time-stamped routine blood biochemistry records reflecting CRP and leucocyte count. The population characteristics were consistent with previous findings.^19,20 Extended analyses were performed to attempt further characterization of the origin of the response as well as the significance its dynamics.

The upper quartile representing the most extensive systemic response at 48 h differed with respect to lower LVEF on CS diagnosis, higher heart rate and lactate, as well as the severity of SCAI grading. Physiological indicators of distress form the basis of CS definition and grading, with perfusion and lactate status continuing to be part of contemporary schemes for shock staging.²¹ Signs of circulatory and end-organ compromise associated with inflammation are supportive of common pathophysiological mechanisms. In terms of comorbidities associated with higher levels of inflammation in the present study, we noted a significantly higher BMI. For the present cohort, DM and plasma glucose have also been associated with 30-day mortality.²²

A trend was observed in lower use of ECMO within the upper quartiles of inflammatory response. The number of cases and level of significance are limited, but the lower prevalence of a treatment modality with increasing severity of illness can either reflect an unmet potential for treatment or be a consequence of survivorship bias, as these patients may not survive to undergo ECMO to the same extent. Findings related to biomarker values which are clinically utilized, such as CRP and leucocyte count as reported herein, may be subject to treatment biases. For instance, patients may undergo further scrutiny with respect to infectious disease diagnostics and treatment as a direct consequence of elevated acute phase response markers. The observed concordance in SCAI grading and ICU duration could indicate that treatment efforts intensify with higher inflammatory status, although causality cannot be inferred.

The association between inflammatory response and mortality persisted regardless of OHCA on presentation. Several factors can be considered in this regard: post-cardiac arrest syndrome may contribute to the inflammatory response following the transient state of universal ischaemia during arrest. Ventilator-assisted pneumonia or other microbial conditions may also contribute to this risk in this group, although we did not note significant differences in the recorded use of mechanical ventilation or bacteraemia and adjusted for it in multivariable models. Our supplementary finding that PCT is associated with mortality is in line with current considerations regarding biomarker-driven efforts to characterize specific inflammatory responses.²³ However, when incorporating PCT into the combined model, the association was diminished. This could support that a separate mechanism is responsible for the PCT response, and that this is not fully explanatory for AMICS patient’s inflammation-driven mortality. Specific microbial causes can be elusive, but ventilator- and cardiopulmonary resuscitation-associated pneumonia have been suggested causes.^16,24

In our analysis of relative CRP change across Days 0–3, a significant increase seemed associated with a lower mortality. Analyses in this context are subject to sampling and survival bias but may give hint to a negative prognostic impact of a diminished ability to manifest a response—or when it has already established on admission. In terms of biomarker dynamics, CRP is often reported to peak no sooner than 48 h, with elevated levels being associated with short-term mortality.^25,26 In extension, the velocity of the CRP change, derived from two measurements within the first 24 h of admission, has been associated with 30-day mortality in a smaller retrospective ST-elevation myocardial infarction (STEMI) study.²⁷ Another study investigated peak CRP within any time—0–96 h of admission and found the discriminatory value for CS severity to be modest compared with PCT and direct measurement of interleukin-6 (IL-6), while 90-day mortality for CRP was non-significant.²⁸ Evidence suggests that the inflammatory component and its prognostic value may be established prior to overt clinical CS; in a large consecutive STEMI cohort, admission leucocyte count was independently associated with 30-day mortality and levels were higher in patients who went on to develop both early and late CS.²⁹ We found that a leucocyte increase across Days 3–7 was associated with a higher mortality, which could reflect a number of factors related to patient’s course of disease, considering the time frame. It does seem plausible that any secondary immune activation implies complications, which could be either of cardiac, infectious, or intrinsic origin.

Contemporary management of CS relies on definition and staging algorithms with emphasis on early intervention towards the underlying cause. Early revascularization is the only strategy with evidence of improved outcomes.³⁰ With escalating severity, intensive care interventions range from pharmacological, haemodynamic optimization to mechanical ventilator, and circulatory support.³¹

As inflammatory pathways are likely involved in the specific pathophysiology of CS, future algorithms could seek to implement inflammatory status in a more direct manner, such as through the addition of SIRS grading or implementation in existing CS staging schemes. In terms of future treatment options, the concept of immunomodulation has received attention for a considerable amount of time and has evolved into a ‘new era’ of anti-inflammatory therapies that have trended towards more biological specificity.^32,33 Randomized controlled trials have investigated IL-6 receptor antagonist treatment in patients that were comatose following OHCA, where intravenous tocilizumab effectively modulated CRP and leucocyte counts at time points within the same range investigated herein.³⁴ The concomitant reduction of troponin T release implied a potential effect on myocardial injury, thus positioning the IL-6 associated signalling pathway cascade for further investigation. In a STEMI-setting, the same agent was shown to increase the myocardial salvage index.³⁵ The absence of increased infection rates and adverse events in the two studies cited is an encouraging safety characteristic, but more research is needed to establish the utility of specific immunomodulatory treatment in AMICS settings.

Strengths and limitations

This validated cohort has the advantage of a large, repeated sample size that allows for estimating biomarker trajectories and association with outcomes in a landmark-type fashion, allowing for a more detailed analysis in subgroups and at sensible time points during the acute phase. The highly dynamic biomarker response coupled with high short-term mortality challenge analyses in retrospective settings, as evident by the ∼50% 48 h mortality. Our findings in terms of admission leucocyte count and mortality could imply that inflammatory responses in the early mortality group may be even higher.

Sampling bias should be considered in these analyses, since blood tests are of clinical origin: more critically ill patients or those perceived as higher risk may undergo more frequent and extensive sampling. Retrospective analyses principally cannot determine causality, meaning it is not possible to firmly conclude whether inflammation is the triggering factor of patient’s higher mortality, alternative to being a composite marker of comorbidity, disease severity, and treatment response.

Limitations concerning the base study populations have previously been discussed.² As has been noted, the national health care infrastructure is considered well organized with regard to centralized acute revascularization, also considering that transport distances are relatively short. Some patients may, however, have been admitted and treated at local hospitals without subsequent transfer to one of the two tertiary centres in the uptake area. Hence, missing cases cannot be entirely excluded, nor can they be readily characterized, although they may be expected to have a higher mortality.

Conclusion

Increased systemic inflammatory response is independently associated with increased mortality in patients surviving cardiogenic shock within the first 48 h, when assessed using combined routine biomarkers.

Ethics

The study was approved by the Danish Patient Safety Authority (3-3013-1133/1) and Danish Data Protection Agency (16/7381).

Supplementary material

Supplementary material is available at European Heart Journal: Acute Cardiovascular Care.

Data availability

Deidentified participant data can be made available upon reasonable request.

References

Author notes