Author + information
- Manuel Almendro-Delia, MD, PhD∗ ( )(, )
- Ivan Nuñez-Gil, MD, PhD and
- Juan C. García-Rubira, MD, PhD
- ↵∗Acute Cardiovascular Care Unit, Cardiology and Cardiovascular Surgery Division, Virgen Macarena University Hospital, Avd. Dr. Fedriani S/N, Sevilla, 41073, Spain
The authors thank our colleagues for their interest in our paper from the RETAKO study (1). In accordance with the findings of Arcari et al. (2), the study observed higher heart rates on admission in subjects with cardiogenic shock (CS) than those without CS (median: 95 [interquartile range (IQR) 80 to 120] beats/min vs. 82 [IQR: 70 to 98] beats/min, respectively; p < 0.0001), but multivariate analysis failed to show any statistically significant risk association (odds ratio [OR]: 1.01 per 10 beats/min; 95% confidence interval [CI]: 0.98 to 1.04). Several reasons may account for these findings in Takotsubo syndrome (TTS).
First, inferring causality is a challenge in observational studies. Regarding TTS, if the outcome of interest (e.g., CS) is present at the time of hospital admission, when predictor variables are measured, ascertaining the time sequence for cause and effect may be difficult. Thus, the temporal sequence was assessed in patients from RETAKO who developed CS beyond 24 h after admission, once predictors were measured, thereby confirming the absence of a significant association between heart rate and CS (OR: 1.33; 95% CI: 0.61 to 2.87). The authors hypothesized that heart rate could be a risk marker rather than a risk factor for CS in TTS, which may reflect both the degree of adrenergic drive and the sum of underlying comorbidities. Therefore, it was argued that the impact of heart rate on mortality in the setting of TTS could be driven by an interaction with low cardiac output syndrome.
Second, the study by Arcari et al. (2) is certainly worthy of consideration because an estimate of causality can be strongly influenced by the manner in which the continuous variable is analyzed. Notwithstanding, the issue of whether the linearity of independent continuous variables and odds for events was either fitted or assumed by default was missed in the study by Arcari et al. (2). It is now widely accepted that a functional form of the relationship between heart rate and outcomes may differ depending upon the clinical picture. Although categorization simplifies the analysis and interpretation of results, it also discards information and can raise critical issues of arbitrariness. It has been suggested that categorization should be avoided in favor of modeling methods based on restricted cubic splines or fractional polynomials, as a way to check whether a linear function is adequate or whether a nonlinear function would substantially improve the fit of data. In this regard, a recent study by the InterTAK (International Takotsubo Registry) investigators addressed the effects of interaction between heart rate and systolic blood pressure (SBP) on all-cause mortality in TTS (3). Using multivariate Cox regression models with flexible smoothing and penalized B-splines, the InterTAK authors found that heart rate on admission was linked to all-cause mortality, showing a significant interaction with SBP (pinteraction = 0.037) (3). Therefore, they demonstrated an association between high heart rate and risk and between low SBP and outcomes. In analyzing RETAKO data, to cope with continuous variables, multivariate fractional polynomials were used instead of B-splines, while maintaining confounding adjustment. Accordingly, when SBP was added to the model, heart rate showed an 18% increase in the risk of long-term mortality (adjusted hazard ratio [HR]: 1.18 for every 10 beats/min increase; 95% CI: 1.04 to 1.33; p = 0.006). In addition, to identify interactions, the predictive ability of the shock index (heart rate-to-SBP ratio), a well-known predictor of morbidity and mortality in emergency scenarios, was explored further. Not surprisingly, shock index was significantly associated with mortality (adjusted HR: 1.30 for every 0.2-U increase; 95% CI: 1.09 to 1.55), especially when high heart rate coexisted with low SBP (Figure 1A).
As stated in the study by Arcari et al. (2), developing tailored treatment approaches based on the underlying factors leading to CS in TTS makes sense. Accordingly, the presence of left ventricular outflow tract obstruction (LVOTO) was the stronger predictor of CS in the present study, regardless of left ventricular ejection fraction and catecholamine response. Prompt identification of LVOTO in unstable patients with suspected TTS may potentially lead to avoidance or discontinuation of catecholamines, preventing further deterioration of the patient's clinical status. Regarding catecholamine use and clinical outcomes in TTS, the present authors agree that catecholamine should be generally regarded as contraindicated in TTS, as further adrenergic activation may worsen prognosis. Accordingly, fine-tuning inotropic support should be the cornerstone of CS management in TTS. Although there is a robust physiological rationale for initiating levosimendan therapy in patients with TTS, a randomized prospective trial is needed to test its safety and efficacy in TTS. In the absence of urgent access to mechanical circulatory support, levosimendan may be preferable to other inotropes in critically ill patients.
Ansari et al. (4) emphasized the risk of catecholamine therapy for treating TTS. In that context, the decision to use catecholamines should be tempered by the consideration that catecholamine could contribute to and exacerbate myocardial toxicity. Unfortunately, data for the type of inotrope were not available in the overall present cohort, but interestingly, a dose-response relationship regarding the duration of inotropic support and mortality (OR: 1.10 per day of treatment; 95% CI: 0.96 to 1.26) was not found. Likewise, the potential association between inotropes and mortality (Figure 1B, model A) was confounded by CS status (Figure 1C, model B). This led to the hypothesis that the clinical scenario may modify the impact of inotropes on mortality. In this regard, the inclusion of CS when adjusting multivariate models was missed in the study by Ansari et al. (4), as up to 86% of patients with CS received catecholamines in that study. Thus, in the present cohort, when both catecholamine therapy and CS where added together to adjust for confounding (model B), use of catecholamines failed to show a significant association with mortality and suffered from collinearity (correlation of regression coefficients: 0.80). Subsequently, how much of an effect on mortality could be attributed to catecholamines themselves, regardless of the presence of a low cardiac output state, remains unknown. Only random treatment assignment could address that issue. Indeed, in the subset of patients without CS, it was found that catecholamine use was not associated with mortality (crude OR: 1.41; 95% CI: 0.32 to 6.11; p = 0.644). Determining causality between catecholamines and mortality in the absence of randomization should be interpreted cautiously, even after multivariate adjustments.
As stated in the study by Ansari et al. (4), further investigation regarding the type and dosages of inotropic support and the clinical setting are deemed necessary in the context of TTS. Meanwhile, we all remain eager for knowledge.
Please note: the authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- 2019 American College of Cardiology Foundation
- Almendro-Delia M.,
- Núñez-Gil I.J.,
- Lobo M.,
- et al.
- Arcari L.,
- Limite L.R.,
- Cacciotti L.,
- et al.
- Böhm M.,
- Cammann V.L.,
- Ghadri J.R.,
- et al.
- Ansari U.,
- El-Battrawy I.,
- Fastner C.,
- et al.