Baseline patients’ characteristics

As represented in Fig. 1, this analysis included 754 CS patients from a total of 49 centers (18 patients excluded for missing Hb level at admission). In accordance with the World Health Organization thresholds, 361 patients (47.9%) presented with anemia at initial admission, including 253 male individuals and 108 female individuals (Hb levels < 13 g/dL and 12 g/dL, respectively). Patients were categorized into four quartiles based on their Hb level at admission: Quartile 1 (Q1) group with a Hb < 11.0 g/dL, Quartile 2 (Q2) group with Hb levels ranging from 11.0 to 12.6 g/dL, Quartile (Q3) group with Hb levels ranging from > 12.6 to 14.0 g/dL, and Quartile (Q4) with Hb > 14.0 g/dL (overall hemoglobin level distribution described as a histogram in Supplementary Fig. 1).

Fig. 1

Flow chart of the study. CS cardiogenic shock

Table 1 provides a comprehensive overview of the baseline characteristics of the study cohort. The mean age was 65.8 (± 14.8) years, with predominance of men (71.8%). Patients in the lowest Hb quartile (Q1) were older (p < 0.01), more frequently male (p < 0.01), and had a lower body mass index (p < 0.01), all demonstrating a significant trend across quartiles. They also exhibited a higher prevalence of comorbidities, including hypertension (p < 0.01), CKD (< 0.01), and active cancer (< 0.01), with a similarly increasing trend observed across quartiles (Ptrend < 0.01). Furthermore, these patients more commonly had a history of cardiomyopathy (increasing from 46.3% to 63.0% between Q4 and Q1, p < 0.01, Ptrend < 0.01), particularly of ischemic and valvular origin, which was associated with a higher rate of previous implantable cardioverter-defibrillator (ICD) (p = 0.02, Ptrend = 0.04)) and more frequent use of beta-blockers and loop diuretics (both with p < 0.01 and Ptrend < 0.01). Based on our adapted classification, 62 patients (8.2%) were categorized as SCAI shock stage B, 257 (34.1%) as SCAI stage C, 328 (43.6%) as SCAI stage D, and 106 (14.1%) as SCAI stage E, with a significant trend towards an increase in the proportion of stage D from Q4 to Q1 (Ptrend = 0.02). AMI was the primary cause of CS in 36.9% of cases and was evenly distributed across quartiles, whereas infectious triggers were more frequent among Q1 patients (decreasing from 19.6% in Q1 to 5.9% in Q4, p < 0.01, Ptrend < 0.01).

Table 1 Baseline characteristics according to hemoglobin level on the first day of managementCS presentation and prognostic markers in the four quartiles of Hemoglobin

As delineated in Table 2, the four quartiles exhibited discernible variations with respect to certain prognostic indicators. Notably, systolic, diastolic, and mean blood pressures were all significantly lower in Q1 (p < 0.01), with a clear trend observed across quartiles (Ptrend < 0.01) for all three measures. They also presented with signs of right heart failure in 57.2% of cases (compared to 43.3% in Q4, p = 0.03, Ptrend < 0.01). From a biological standpoint, a lower Hb level was associated with higher levels of creatinine and natriuretic peptides (both NT-proBNP and BNP), as well as a lower prothrombin time, all demonstrating a highly significant trend (Ptrend < 0.01). No significant difference was found for skin mottling, signs of left heart failure, and medical history of initial cardiac arrest.

Table 2 Clinical, echocardiographic, and laboratory parameters according to hemoglobin level on the first day of managementIn-hospital management according to the four quartiles of hemoglobin

Data on in-hospital management are presented in Table 3. Notably, patients in the lowest Hb quartile (Q1) were administered norepinephrine significantly more frequently than those in Q4 (63.0% vs. 47.6%, p = 0.02, Ptrend < 0.01). Additionally, patients in Q1 also required RRT significantly more often (26.5% in Q1 vs. 11.2% in Q4, p < 0.01, Ptrend < 0.01) and had a markedly higher need for RBC transfusion support (31.2% in Q1 vs. 11.2% in Q4, p < 0.01, Ptrend < 0.01). Lastly, a decreasing trend was observed in the use of microaxial flow pumps across quartiles with lower Hb levels (Ptrend < 0.01).

Table 3 In-hospital management according to hemoglobin level on the first day of managementCS evolution according to quartiles of hemoglobin

First, whether at 1 month (Ptrend = 0.035) or 1 year (Ptrend < 0.01), a strong significant trend towards increased all-cause mortality was observed across Hb quartiles, suggesting a gradual correlation between quartiles of Hb level and all-cause mortality (Fig. 2). Specifically, compared with the Q4 group (taken as reference), Q1 patients demonstrated a 1.64-fold higher mortality rate at one month (95% CI: 1.09–2.47, p = 0.02), which amplified at one year, with a 2.53-fold increase in events (95% CI: 1.84–3.49, p < 0.01). Patients from the Q2 group also showed increased mortality at 1 month (HR 1.65 [95% CI 1.10–2.48], p = 0.02), which was also further confirmed and amplified at 1 year (HR 2.01, 95% CI 1.44–2.79), p < 0.01). In a multivariate Cox regression model (adjusted for baseline characteristics, left ventricular ejection fraction [LVEF], arterial blood lactate levels, CS triggers, aMCS, acute renal replacement therapy, invasive respiratory support, and arterial oxygen pressure), low hemoglobin levels emerged as one of the few significant cofactors associated with all-cause mortality, both at 1 month (aHR 1.76 [95% CI 1.02–3.04], p = 0.04 and 1.72 [95% CI 1.03–2.87], p = 0.04 for Q1 and Q2, respectively, compared with Q4) and at 1 year (aHR 2.13 [95% CI 1.40–3.24], p < 0.01 and 1.83 [95% CI 1.22–2.73], p < 0.01 for Q1 and Q2, respectively, compared with Q4) (Table 4). The sensitivity analysis performed through multivariate logistic regression yielded similar results, with significantly increased 30-day mortality for Q1 (aOR 1.11 [1.01–1.23], p = 0.04) and Q2 (aOR 1.12 [1.02–1.25], p = 0.02) as well as 1-year mortality for Q1 (aOR 1.24 [1.10–1.39], p < 0.01) and Q2 (aOR 1.18 [1.06–1.32], p < 0.01) (Supplementary Table 1).

Fig. 2

Short- and long-term mortality outcomes after CS according to baseline hemoglobin level. A represents 1-month overall mortality. B focuses on 1-year mortality. The cumulative incidences of 1-year and 1-month mortality were estimated with the use of the Kaplan–Meier method; hazard ratios and 95% confidence intervals were estimated with the use of Cox regression models. Patients were categorized into four quartiles based on their hemoglobin level at admission: Quartile 1 group with a Hb < 11.0 g/dL, Quartile 2 group with Hb levels ranging from 11.0 to 12.6 g/dL, Quartile group with Hb levels ranging from > 12.6 to 14.0 g/dL, and Quartile with Hb > 14.0 g/dL. CS cardiogenic shock, HR hazard ratio

Table 4 Description of all hazard ratios and 95% confidence intervals of all variables tested in multivariate analysis models for 1-month and 1-year all-cause mortalityRestricted cubic spline curves

Monotonic relationships between Hb level and all-cause mortality were confirmed by spline analysis results, with P-overall at 0.04 and < 0.01 for 1-month and 1-year mortality (Fig. 3). The optimal cut-off for Hb’s impact on 1-month mortality was 12.6 g/dL and 12.7 g/dL for 1-year mortality. Besides, the p-value for non-linearity was higher than 0.05, suggesting a possible linear association between Hb and mortality.

Fig. 3

Restricted cubic spline curves for the relationship between hemoglobin and all-cause mortality. A represents the restricted cubic spline curve for 1-month mortality. B represents the restricted cubic spline curve for 1-year mortality. The gray area represents the 95% confidence interval

Subgroups analysis

First, we observed that the relationship between Hb levels and mortality in the overall population does not seem to apply specifically to women, either at 1 month (Ptrend = 0.9) or 1 year (Ptrend = 0.09), with no quartile showing a significant association. In contrast, this relationship is similarly expressed in men, with a 2.02-fold higher mortality rate at 1 month (95% CI 1.25–3.26, p < 0.01) and a 2.84-fold increase in events at 1 year (95% CI 1.94–4.14, p < 0.01) (Supplementary Fig. 2). Moreover, no significant trend was found between Hb levels and mortality for patients with CKD (Ptrend = 0.53 at one month and 0.43 at one year), with no quartile reaching statistical significance. However, the link between anemia and mortality was again highlighted in non-CKD patients, with higher mortality observed in the Q1 group at both 1 month (HR 1.69 [95% CI 1.06–2.70], p = 0.03) and 1 year (HR 2.47 [95% CI 1.72–3.57], p < 0.01) (Supplementary Fig. 3). Besides, a lower Hb level was still associated with increased mortality for AMI-CS patients at 1 year (P trend < 0.01, HR 2.17 [1.32–3.56], p < 0.01), but not at 1 month (P trend = 0.57, with no quartile being significant). In contrast, a strong negative impact of anemia on mortality was observed for NICS patients, with a 1-month HR ranging from 2.10 to 2.86 for the first three quartiles (P trend < 0.01), which persisted at 1 year for Q1 (HR 2.59 [1.72–3.91], p < 0.01) and Q2 (HR 2.65 [1.76–3.99], p < 0.01) (Supplementary Fig. 4). Eventually, the detrimental effect of low hemoglobin levels was also observed in patients aged < 67 years, at both 1-month (HR for Q1: 1.90 [1.05–3.46], p = 0.04, Ptrend = 0.03) and 1-year (HR for Q1: 2.94 [1.85–4.66], p < 0.01, Ptrend < 0.01). In contrast, for patients aged ≥ 67 years, the negative effect of anemia was only observed at 1 year (Supplementary Fig. 5).