FormalPara Take-home message

In this study, two in five acutely admitted adult ICU patients had thrombocytopenia; half of those had thrombocytopenia at ICU admission, which was associated with increased mortality. Platelet transfusions were frequently used in patients with severe and very severe thrombocytopenia, and most transfusions were prophylactic.

Introduction

Thrombocytopenia, usually defined as a platelet count < 150 × 109/L, occurs frequently in intensive care unit (ICU) patients [1, 2]. It is often multifactorial with causes including conditions and interventions resulting in reduced platelet production or increased platelet consumption, destruction, or sequestration, and haemodilution [3]. The prevalence and cumulative incidence have been reported in ranges from 8% to 77% and 8% to 56%, respectively, depending on case mix and definitions [4, 5]. Observational studies suggest that thrombocytopenia may impact ICU patients’ short- and long-term outcomes, including increased bleeding, transfusion requirements, use of life-support, length of ICU- and hospital stay, and mortality [4, 5]. However, uncertainty exists as many studies have been single-centred with low external validity [1, 2, 6,7,8,9].

Thrombocytopenia may prompt ICU physicians to delay or withhold invasive procedures, reduce thrombosis prophylaxis and consider platelet transfusions [10]. Prophylactic platelet transfusions are frequently used in thrombocytopenic ICU patients to reduce the risk of bleeding [11, 12], but evidence from randomised trials in ICU settings is sparse [13, 14], limited to a recent study evaluating platelet transfusion before central venous catheter (CVC) placement [15].

We conducted an international, prospective, inception cohort study to provide contemporary epidemiological data describing the frequency of thrombocytopenia in ICU patients, its risk factors, current platelet transfusion practice, and clinical outcomes including mortality. We hypothesised that thrombocytopenia would be frequent and associated with worse outcomes and that specific risk factors for thrombocytopenia exist.

Methods

Study design and setting

This was an international, prospective, inception cohort study conducted and analysed in compliance with a published protocol and statistical analysis plan with minor modifications (electronic supplementary material (ESM) 1) [16]. The manuscript was prepared according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement [17] (ESM 2).

The study was performed in 52 ICUs across 10 countries in Europe and North America (ESM 3 and 4) and overseen by a steering committee (ESM 5). Study ICUs were invited to participate through the research network Caring for critically ill immunocompromised patients: Multinational Network (Nine-I).

The study was approved by the Danish Centre for Regional Development (R-21012287) and registered at the Capital Region’s Centre for Data Compliance (P-2021-262). We obtained all necessary approvals in the other participating countries and collected informed consent from patients and/or surrogates if required.

Inception periods of 14 consecutive days were conveniently chosen by participating ICUs from May 2021 to July 2022. Included patients were followed for a maximum of 90 days.

Population

During inception periods, all consecutive adults (≥ 18 years) who were acutely admitted to the ICU were screened for inclusion. Patients transferred from other ICUs were considered eligible if they otherwise met the inclusion criteria. We excluded patients with elective open-heart surgery during current hospitalisation and those who had previously been enrolled or declined informed consent.

Data collection and management

We developed an electronic case report form (eCRF) using the Research Electronic Data Capture (REDCap) tool hosted by the Capital Region of Denmark [18, 19]. The eCRF was pre-tested at the coordinating site and local investigators had access to a training version before study initiation.

For each study ICU, we collected data on type of hospital and ICU, number of ICU beds and presence of local guidelines for platelet transfusions (ESM 6).

Patient data were collected from medical records as described in the protocol [16]. Baseline data collected at ICU admission included demographics, comorbidities, selected treatments, Simplified Mortality Score for the Intensive Care Unit (SMS-ICU, an illness severity score ranging from 0 to 42 points with higher scores indicating higher predicted 90-day mortality) (ESM 7) [20, 21], platelet count, and presence of bleeding (ESM 8).

Daily data were collected during ICU stay and resumed upon readmission or transfer to another study ICU for a maximum of 90 days. We collected data on the use of life-support [vasopressor/inotropes, mechanical ventilation, renal replacement therapy, extracorporeal membrane oxygenation (ECMO)], selected treatments, platelet counts, bleeding events, thrombotic events, and transfusions (ESM 9). For platelet transfusions used within ICU departments, we collected data on indications for transfusion (pre-procedural: covering invasive procedures or surgery; prophylactic: reducing the risk of bleeding; therapeutic: treating bleeding) and platelet count before transfusion (ESM 1 and 9). On day 90, we assessed the vital status and number of days in the hospital (ESM 10).

Definitions and classifications

Thrombocytopenia

We defined thrombocytopenia as a platelet count < 150 × 109/L and classified thrombocytopenia into three categories: ‘baseline thrombocytopenia’ defined as patients with thrombocytopenia at ICU admission, ‘ICU thrombocytopenia’ defined as patients with thrombocytopenia during ICU stay without thrombocytopenia at ICU admission and ‘any thrombocytopenia’ defined as patients who had thrombocytopenia at ICU admission and/or during ICU stay. Severity of thrombocytopenia was classified as mild (100–149 × 109/L), moderate (50–99 × 109/L), severe (20–49 × 109/L), and very severe (< 20 × 109/L), based on the platelet count at ICU admission and the nadir platelet count during ICU stay and overall, for baseline-, ICU-, and any thrombocytopenia, respectively [16].

Bleeding

We defined and graded bleeding according to a modified World Health Organization (WHO) classification [22,23,24,25]. Grade 1 and 2 covers bleedings that do not require red blood cell (RBC) transfusion. Grade 3 covers bleedings requiring up to two RBC transfusions and central nervous system bleedings without symptoms. Grade 4 covers bleedings requiring either (1) more than two RBC transfusions, (2) intubation and mechanical ventilation or (3) surgical intervention, and bleedings from critical sites with symptoms and fatal bleedings (ESM 11) [16].

Outcomes

The primary outcome was the number of patients with any thrombocytopenia. Secondary outcomes were numbers of patients with mild, moderate, severe, and very severe thrombocytopenia; 90-day mortality; days alive without use of life-support; days alive and out of hospital; numbers of patients with at least one bleeding event, at least one thrombotic event, and at least one platelet transfusion during ICU stay; number of platelet transfusions and volumes transfused per patient; and number of transfusions for each indication (ESM 12) [16].

Process variables

Process variables included daily platelet count before in-ICU platelet transfusions for each indication and number of RBC transfusions (ESM 1 and 12).

Statistical methods

Sample size

As we planned to include patients regardless of illness severity and expected length of ICU stay, we made a conservative assumption, that the frequency of any thrombocytopenia would be 20% [4, 5]. We aimed to enroll a minimum of 1,000 patients to determine the frequency with a 95% confidence interval (CI) of 18–23% [16].

Descriptive data

Analyses were performed using R (version 4.2.0) following the statistical analysis plan with minor modifications (ESM 1) [16]. Categorical variables were reported as numbers and percentages and continuous variables as medians with interquartile ranges (IQRs). Primary and secondary outcomes were reported descriptively with 95% CIs calculated using the Clopper–Pearson method for categorical variables and non-parametric percentile-based bootstrapping with 100,000 samples for continuous variables. Results were stratified on the three categories of thrombocytopenia, and for the secondary outcomes, also on the severity of thrombocytopenia [16].

Regression analyses

We used pre-specified logistic regression models [16] to assess a set of potential baseline risk factors for ICU thrombocytopenia (< 150 × 109/L and < 50 × 109/L) selected a priori based on literature review [4, 5], clinical knowledge, and the expected number of events. We assessed sex, haematological malignancy, immune deficiency, acute or chronic liver failure, SMS-ICU, bleeding and septic shock in unadjusted models and in a model adjusted for all selected risk factors [16]. We modified the planned definition of ‘immune deficiency’ to, ‘non-acquired immune deficiency syndrome (AIDS)-, non-cancer-related immune deficiency’ to avoid any unintended overlap with cancer patients (ESM 1). Patients with baseline thrombocytopenia < 150 × 109/L were excluded from these analyses.

We assessed associations between baseline thrombocytopenia (< 150 × 109/L and < 50 × 109/L) and 90-day mortality in pre-specified logistic regression models adjusted for potential confounders including sex, country, haematological malignancy, SMS-ICU, and septic shock selected based on literature review [4, 5], clinical knowledge and expected number of events.

We used Nagelkerke’s R2 [26] and calibration plots [27] to assess model fits (ESM 24). We conducted additional sensitivity analyses including complete case analyses and Cox proportional hazards regressions to assess the impact of missing platelet counts and time-to-event and competing risks, respectively (details in ESM 25 and 27).

Missing data

We handled missing platelet count data, as planned, using logical imputation, and assumed that missing baseline platelet counts were equal to the first platelet count in ICU (ESM 13). Twenty-eight patients were transferred directly from a study ICU to a non-study ICU where daily variables (clinical events, biochemistry, etc.) could not be obtained. We assumed that no events occurred during days spent in non-study ICUs. Following these logical imputations, we performed complete case analysis due to limited remaining missing data (< 5% of patients) [16].

Results

We included 1168 ICU patients from 52 ICUs across 43 centres in 10 countries. Two patients withdrew consent and were excluded (ESM 14). Most study ICUs were mixed ICUs (76.9%) in university hospitals (94.2%). Few (15.4%) had ICU-specific guidelines for platelet transfusions (ESM 15).

Baseline characteristics

Patients with thrombocytopenia were more often males, admitted from hospital wards, and were more severely ill than those without thrombocytopenia (Table 1 and ESM 16). More patients with thrombocytopenia had haematological malignancy, non-AIDS-, non-cancer-related immune deficiencies, acute and chronic liver failure, septic shock, and bleeding at baseline and more received treatment with haematopoietic stem cell transplantation, chemotherapy and platelet transfusions compared to those without thrombocytopenia.

Table 1 Baseline characteristics
Full size table

Common characteristics in patients with baseline thrombocytopenia, were admission from hospital wards, haematological malignancy, and treatment with haematopoietic stem cell transplantation, chemotherapy, and platelet transfusions prior to ICU admission while surgery before ICU admission and admission from emergency settings, operating or recovery rooms were more common in patients with ICU thrombocytopenia.

Primary outcome

The frequency of any thrombocytopenia was 43.2% (95% CI 40.4–46.1); 23.4% (21–26) had baseline thrombocytopenia, and an additional 19.8% (17.6–22.2) developed ICU thrombocytopenia.

Secondary outcomes

ICU thrombocytopenia occurred on median (IQR) day 2 (1–3) with few occurrences after day 5 (10.8%) from ICU admission (ESM 17) and most patients with baseline thrombocytopenia also had thrombocytopenia during ICU stay (95.6%). Among patients with any thrombocytopenia, 42.5% (95% CI 38.1–46.9) had mild-, 31% (26.9–35.2) moderate-, 12.9% (10.1–16.1) severe- and 13.7% (10.8–17) very severe thrombocytopenia (ESM 18) with nadir platelet counts occurring on median (interquartile range, IQR) 3 (1–4) days from ICU admission. Nadir platelet counts for patients with baseline- and ICU thrombocytopenia occurred on median (IQR) days 2 (0–4) and 4 (2–5) from ICU admission, respectively (day 0 representing baseline). The trajectories of thrombocytopenia from baseline through ICU stay are presented in Fig. 1 and ESM 19.

Fig. 1
figure 1

The trajectories of thrombocytopenia. Trajectories of thrombocytopenia in all patients (A) and exclusively in those who had thrombocytopenia (B). The severity of baseline- and ICU thrombocytopenia was based on the platelet count at baseline and the nadir platelet during ICU stay, respectively. The flows are coloured according to the baseline severity. Patients whose thrombocytopenia remained unchanged (i.e., unchanged severity) from baseline throughout their ICU stay are represented by flows moving between identical colours. Patients whose thrombocytopenia got better or worse (i.e., decreased or increased severity) during their ICU stay compared to their starting point at baseline are represented by flows moving between different colours. Only 22 patients had less severe thrombocytopenia during their ICU stay compared to their starting point at baseline (represented by flows mowing from “warmer” to “colder” colour tones)

Full size image

The overall 90-day mortality was 25.9% (95% CI 23.4–28.5) (ESM 20) and differed between patients with- and without thrombocytopenia at 34.6% and 19.3%, respectively. The highest mortality was observed in patients with severe- and very severe thrombocytopenia (Table 2).

Table 2 Secondary clinical outcomes
Full size table

Overall, the median number of days alive without use of life-support was 87 (IQR 0–90) and days alive and out of hospital was 63 (0–79). Patients with thrombocytopenia had fewer days alive without use of life-support and fewer days alive out of hospital than those without thrombocytopenia (Table 2). More patients with thrombocytopenia received vasopressors (73% vs 45.8%), invasive mechanical ventilation (63.3% vs 44.6%), RRT (25% vs 5.3%) and ECMO (2.8% vs 0.6%) compared with those without thrombocytopenia (ESM 21).

In total, 16% of patients bled in ICU (95% CI 14–18.3) (ESM 20). Among patients without thrombocytopenia, 7.3% bled in ICU, most of whom (81.2%) had bleeding of grade 1 or 2. In contrast, 27.6% of patients with thrombocytopenia bled in ICU, of whom 57.6% had bleeding of grade 3 or 4. Proportionally more patients with severe- and very severe thrombocytopenia had grade 3 or 4 bleeding compared to patients with no- or less severe thrombocytopenia (Table 2). Thrombotic events occurred in 5.8% (95% CI 4.6–7.3) of all patients (Table 2 and ESM 20).

Platelet transfusions were almost exclusively used in patients with thrombocytopenia, of whom 22.6% (95% CI 19–26.5%) were transfused (Table 2 and ESM 20). Patients transfused with platelets received median (IQR) 3 (1–5) platelet transfusions. Most platelet transfusions (90.9%) took place within ICU departments (in-ICU) and fewer in operating theatres (9.1%) (EMS 22). The number of in-ICU transfusions increased with higher severity of thrombocytopenia (Fig. 2) and most in-ICU transfusions were prophylactic (64.3%) followed by therapeutic- (27.6%) and pre-procedural transfusions (8.1%) with median platelet counts before transfusion of 15 × 109/L (IQR 9–26), 40 × 109/L (23–76) and 31 × 109/L (20–54), respectively. Platelet counts before transfusion for specific procedures are presented in ESM 23. More patients with thrombocytopenia received RBC transfusion than patients without thrombocytopenia (Table 2 and ESM 20).

Fig. 2
figure 2

In-ICU platelet transfusions and indications. Number of in-ICU platelet transfusions used in patients with- and without thrombocytopenia. Patients with severe- and very severe thrombocytopenia received 93% of all in-ICU platelet transfusions and in patients with very severe thrombocytopenia, 76% of in-ICU transfusions were prophylactic. We did not collect data on indications for 57 platelet transfusions used during surgery, thus these are excluded (see ESM 22 for details)

Full size image

Risk factors for ICU thrombocytopenia

In unadjusted and adjusted analyses, male sex, non-AIDS-, non-cancer-related immune deficiency, acute or chronic liver failure, higher illness severity [20, 21], bleeding, and septic shock at baseline were associated with ICU thrombocytopenia < 150 × 109/L. Meaningful interpretation of risk factors for ICU thrombocytopenia < 50 × 109/L were hampered by few events (n = 26) (Table 3). Results were largely consistent in sensitivity analyses (ESM 25).

Table 3 Baseline risk factors for ICU thrombocytopenia
Full size table

Baseline thrombocytopenia and mortality

When adjusting for country, sex, haematological malignancy, SMS-ICU and septic shock, baseline thrombocytopenia < 150 × 109/L was associated with increased 90-day mortality (OR 1.7, 95% CI 1.19–2.42), while some uncertainty remained about the effect of baseline thrombocytopenia < 50 × 109/L (1.79, 95% CI 0.89–3.61) (ESM 26). Sensitivity analyses yielded largely consistent results (ESM 27).

Discussion

In this international prospective cohort study of acutely admitted adult ICU patients, we found that 43% had thrombocytopenia; 23% at ICU admission and 20% developed it during ICU stay. Patients with thrombocytopenia had worse outcomes, including 90-day mortality, and several conditions increased the risk of developing thrombocytopenia during ICU stay. Twenty-three percent of patients with thrombocytopenia received platelet transfusion and most in-ICU transfusions were prophylactic.

The observed frequency of thrombocytopenia was similar to findings in smaller cohort studies in general mixed ICU populations [2, 28], and in a larger cohort of ICU patients receiving thromboprophylaxis [29]. However, variation exists, which is likely due to differences in case mix, illness severity, follow-up time and definitions of thrombocytopenia [1, 9, 30, 31].

We found that male sex, non-AIDS-, non-cancer-related immune deficiency, acute or chronic liver failure, higher illness severity, bleeding, and septic shock at baseline increased the risk of developing ICU thrombocytopenia. This aligns well with previously published data [29, 30, 32], although one study using a different definition of thrombocytopenia (< 100 × 109/L) found female sex to be associated with the development of ICU thrombocytopenia [30]. Comparing risk factors across studies is challenging as case mix, definitions of thrombocytopenia and analytic methods differ [29, 30, 32].

Consistent with previous studies [4, 5, 29, 30], we observed that patients with thrombocytopenia had worse outcomes, including higher rates of bleeding, transfusion and mortality, fewer days alive without use of life-support and days alive and out of hospital. Associations between thrombocytopenia and bleeding, transfusion, and mortality followed a “dose–response” pattern with worse outcomes accompanying more severe thrombocytopenia. Patients with severe- and very severe thrombocytopenia had markedly higher mortality but were also more likely to experience grade 3 or 4 bleeding and to receive transfusions compared with patients with no- or less severe thrombocytopenia. In this heterogeneous ICU population, it remains unclear whether the observed excess mortality is directly related to thrombocytopenia and its consequences, such as bleeding and transfusions, or whether thrombocytopenia merely represents a marker of illness severity or the pathophysiology of the underlying disease. Importantly, identified risk factors and association with mortality do not represent causal relationships.

Most in-ICU platelet transfusions were prophylactic (64%) with a median platelet count before transfusion of 15 × 109/L. Lower proportions of prophylactic platelet transfusions (40–45%) with higher median platelet counts before transfusion have been reported, which likely reflects differences in case mix and changes in practice over time [11, 12]. The European Society of Intensive Care Medicine (ESICM) recommends against prophylactic platelet transfusion unless the platelet count falls below 10 × 109/L, but the recommendation is only conditional due to very low certainty of evidence and randomised trials evaluating this strategy in general ICU patients are warranted [13].

In our cohort, prophylactic pre-procedural platelet transfusions constituted 8% of in-ICU platelet transfusions and currently, ESICM makes no recommendations regarding this strategy for patients with platelet counts between 10–50 × 109/L [13]. A recent non-inferiority trial enrolling patients in haematology wards and ICUs with platelet counts between 10–50 × 109/L showed that withholding versus providing a single prophylactic platelet transfusion before CVC placement was not non-inferior and resulted in more CVC-related bleedings [15]. Interestingly, the occurrence of bleeding was lower in ICU patients compared with haematology ward patients, which might support a no-transfusion strategy with intensive monitoring in ICU patients [15].

This study has some strengths. It was prospective and conducted, analysed and reported in compliance with a published protocol [16]. Further, contrary to many previous studies [1, 2, 6,7,8,9], we included a large, unselected, multinational ICU population. This study also provides data on the use of in-ICU platelet transfusions across several countries, which complements previous national studies [11, 12, 33].

Our study also has some limitations. First, 17% of the patients did not have a baseline platelet count and although we handled this as planned, by using the first available platelet count in the ICU measured on median (IQR) day 1 (1–1), some misclassification cannot be ruled out. Second, 7% of patients were transferred directly from other ICUs, with a median stay of 3 days before inclusion. This could underestimate event rates and overestimate baseline thrombocytopenia at the expense of ICU thrombocytopenia. Third, patients with haematological malignancies constituted 8% of the population and 15% of the cases with thrombocytopenia; frequencies of thrombocytopenia may be lower in ICU populations with lower proportions of patients with haematological malignancies. Fourth, 2% of patients were transferred to a non-study ICU without available daily variables; these were analysed under the assumption that no events occurred during days spent in the non-study ICU. Fourth, descriptions of platelet counts before platelet transfusions were added post hoc and assumptions made during computation may have affected results. Fifth, we used pre-specified logistic regression analyses to assess risk factors for ICU thrombocytopenia and associations between baseline thrombocytopenia and 90-day mortality. Although post hoc sensitivity analyses including Cox proportional hazard models accounting for time-to-event and competing risks produced similar results, our regression analyses should be interpreted carefully considering the limited number of included variables and risk of residual confounding (e.g., the effects of baseline thrombocytopenia on 90-day mortality may vary according to the underlying cause, which was not assessed in this study). Lastly, meaningful interpretation of risk factor analyses for ICU thrombocytopenia < 50 × 109/L was hampered by few events and the analysis of baseline thrombocytopenia < 50 × 109/L and 90-day mortality was likely underpowered.

Conclusions

In this international, prospective cohort study, 43% of adult acutely admitted ICU patients had thrombocytopenia. Patients with thrombocytopenia had worse outcomes including increased 90-day mortality. Baseline risk factors for the development of thrombocytopenia during ICU stay included male sex, non-AIDS-, non-cancer related immune deficiency, liver failure, higher illness severity, bleeding, and septic shock. Platelet transfusions were given to 23% of patients with thrombocytopenia and most were prophylactic.