Introduction

Thrombocytopenia (TP) is a common laboratory finding in intensive care unit (ICU) patients. A 150 × 109/L threshold is used in most studies to define TP [1–11] but its definition varies and some authors use other cut-offs such as 100 × 109/L [12–17], 50 × 109/L [18, 19], and 30 × 109/L [20]. In addition, most studies do not include a repeat platelet count to confirm TP, thereby limiting the accuracy of TP diagnosis [5, 6, 18]. Depending on the definition of TP, the clinical setting, and the case mix population, the frequency ranges from 8.3 to 67.6 % for prevalent TP (TP diagnosed on ICU admission) and from 13.0 to 44.1 % for incident TP (TP acquired during ICU stay) [21]. The causes of TP in critically ill patients are usually multifactorial and mainly result from enhanced platelet consumption, reduced platelet production, hemodilution, or a combination of any of these mechanisms. However, to our knowledge, very few studies have extensively assessed the causes of TP [6, 16].

Several works have addressed the consequences of TP as a determinant of severe bleeding [3, 5, 6, 18] but with conflicting results [3, 18]. Platelet count nadir and kinetics have been reported to be associated with adverse outcome in ICU patients [3, 4, 22–24]. The consequences of TP in critically ill patients are uncertain and hence its management remains unresolved.

We decided, therefore, to assess the epidemiology of ICU thrombocytopenic patients. We focused on TP etiologies by performing an extensive diagnostic workup that included bone marrow aspiration.

Patients and methods

Study population

This prospective observational study was carried out between July 2005 and November 2006. One medical and two mixed ICUs in a university hospital and two mixed ICUs in general hospitals took part. Patients older than 18 years who presented with TP either on admission or during ICU stay were included. Exclusion criteria included pregnancy and an ICU length of stay less than 48 h. In all patients, platelet counts were performed daily from ICU admission to ICU discharge. The study protocol was approved by the institutional ethics committee. Informed consent was required from the patients or their next of kin.

Study protocol

The day of inclusion was the day of TP diagnosis. On inclusion, the following tests were performed and their results collected: activated partial thromboplastin time (aPTT), prothrombin time (PT), expressed as a percentage of the ratio of the PT in patients over the PT in controls (STA NEOPLASTINE CI PLUS; Diagnostica Stago, France), plasma level measurement of D-dimer (VIDAS D-dimer, BioMerieux, France), fibrinogen (STA FIBRINOGEN; Diagnostica Stago, France), serum level measurement of creatinine, ferritin, triglyceride, bilirubin, haptoglobin, lactate dehydrogenase (LDH), aspartate aminotransferase (ASAT), alanine aminotransferase (ALAT), vitamin B12, folic acid, and antibodies against Epstein–Barr virus (EBV) cytomegalovirus (CMV), human immunodeficiency virus (HIV), and B and C hepatitis using enzyme-linked immunosorbent assay (ELISA) techniques. A platelet antibody test was performed in patients who had had platelet transfusion administered within the 3 weeks before inclusion. In patients with previous heparin exposure and an intermediate or high suspicion of heparin-induced TP (HIT) as defined by a 4-T score greater than 3 [25], a PF4/heparin ELISA was carried out and when PF4/heparin antibodies were detected, a heparin-induced platelet activation test was performed. Bone marrow aspiration was performed and all aspirates were reviewed by an experienced hematologist. Daily platelet counts from inclusion to 14 days after inclusion or until ICU discharge if patients were discharged earlier were also recorded. The other data collected are given in Supplemental File 1.

Definitions

Absolute TP was defined by a platelet count below 100 × 109/L [12, 17] and relative TP by a decrease in the platelet count of more than 30 % [6] as compared with the highest values of the platelet counts observed within the preceding 4 days in patients without absolute TP. TP was confirmed using a subsequent platelet count control performed within the following 24 h. Patients with relative TP who developed absolute TP during their ICU stays were allocated to the absolute TP group.

A serious bleeding event (SBE) was defined by at least one of the following criteria: central nervous system hemorrhage, blood loss requiring the transfusion of three or more units of red blood cells (RBCs) in a 48-h period, bleeding requiring surgery or endoscopy, or fatal bleeding [26]. All other bleeding events were considered to be of minor importance.

Other definitions used are given in Supplemental File 2 [27–33].

Statistical analysis

Data were entered onto a computer and analyzed using Epi-info version 6 (Center for Disease Control, Atlanta, GA, USA) and Statview 5.0 statistical packages (SAS Institute Inc., SAS Campus Drive, Cary, NC 27515, USA). Values are expressed as mean ± SD or median and IQR or as a percentage of the group from which they were derived (categorical variables). Categorical data were compared using the χ 2 test or Fisher exact test, as appropriate. Continuous variables were compared using the Student’s t test or Mann–Whitney U test when appropriate. The nadir platelet count taken into account in the models was the lowest platelet count during ICU stay in patients without severe bleeding events and the lowest platelet count prior to the occurrence of SBE in the others. In patients who were administered prophylactic platelet transfusion, the nadir platelet count value entered in the model was the platelet count at the time of the first platelet transfusion. The risk of the first SBE was assessed in patients who did not receive prophylactic platelet transfusion by a logistic regression model. The relationships between hospital mortality and the study variables were assessed using a logistic regression model in which vital status at hospital discharge was the variable of interest. Variables with P < 0.20 by univariate analysis were entered into the logistic regression model, where they were kept if the P value in the multivariable context was less than 0.05 (stepwise variable-selection method). All statistical tests were two-tailed, and P values less than 0.05 were considered to be significant.

Results

Population

Of the 301 patients who met the inclusion criteria, 208 were diagnosed with absolute TP and 93 with relative TP (Supplemental File 3). The characteristics of the population are shown in Table 1. A total of 208 patients presented with absolute TP diagnosed either on admission (N = 102) or during ICU stay (N = 106). In the latter 106 patients, absolute TP developed at a median time of 3 days (IQR = 2–4) after ICU admission. Relative TP was ultimately diagnosed in 93 patients and developed at a median time of 4 days (IQR = 3–5) after ICU admission. The median duration of TP during the first 14 days following TP diagnosis was 4 days (IQR = 2–7) in patients with absolute TP. The daily platelet counts in the 14 days following TP diagnosis in patients with absolute and relative TP are given in Supplemental File 4.

Table 1 Characteristics of the study population
Full size table

Medical conditions associated with TP

The median number of TP etiologies per patient was two (Table 2). Multiple causes of TP were more frequently observed in patients with absolute TP. Infections (sepsis/severe sepsis and septic shock) were the leading medical condition associated with TP. The sites of infection and etiologic microorganisms are shown in Supplemental File 5. Viral diagnostic workup was performed in 261 patients (87 %) and identified 69 viral infection-associated TP in 62 (24 %), due to EBV in 33 cases, CMV in 21, hepatitis B in 5, hepatitis C in 7, and HIV in 3. In 2 patients viral infection was the only medical condition associated with TP. In the 123 patients with DIC, infection was the main cause in 92 (75 %), followed by massive transfusion in 17 (7 %) and cancer in 6 (5 %). Of the 32 patients with drug-induced TP, 14 (44 %) had no other identified cause. In 169/301 patients (56 %), prior heparin exposure was present before TP diagnosis, including unfractionated heparin in 87 patients and low molecular weight heparin in 82. A diagnostic procedure for HIT was performed in 70 patients (41 %): heparin-PF4 immunoassay in all patients, and functional assay in 20 patients. The diagnosis of HIT was ultimately considered in 5/169 patients (3 %). Hemophagocytic syndrome was diagnosed in 14 patients, of whom 5 had at least five diagnostic criteria and the other 9 four criteria.

Table 2 Etiologies of thrombocytopenia
Full size table

Bone marrow aspiration

Bone marrow samples were taken in 255 patients and could be analyzed in 238. No complications were observed after aspiration. Decreased platelet production was observed in 17 cases (7 %). A total of 102 pathologic findings were identified in 44 % (71/163) of patients with absolute TP and in 28 % (21/75) of patients with relative TP (P = 0.03). In 54 patients, cytological signs like those classically described in patients with vitamin B12 or folate deficiencies were observed on bone marrow aspirates but were associated with serum vitamin deficiency in only 25. There was no relationship between vitamin deficiency and the presence of cytological signs (P = 0.22). However, megaloblastosis was more frequently observed in patients with than without vitamin deficiency (P = 0.02). In 52 (22 %) of the 238 patients with analyzable results, bone marrow aspirates yielded a positive diagnosis not previously established. Results of bone marrow aspirates yielding a previously unestablished diagnosis were observed in 24/96 patients (25 %) with DIC and in 28/142 (20 %) without (P = 0.40), in 39/182 patients (21 %) with sepsis and/or septic shock and in 13/56 (23 %) without (P = 0.9), and in 40/163 patients (25 %) with absolute TP and 12/75 (16 %) with relative TP (P = 0.19). A diagnosis established on the basis of bone marrow results and susceptible to having a significant impact on subsequent patient management was observed in 24 cases (11 %). Results of bone marrow aspirates susceptible to having a significant impact on subsequent patient management were observed in 13/96 patients (14 %) with DIC and in 11/142 (8 %) without (P = 0.22), in 17/182 patients (9 %) with sepsis and/or septic shock and in 7/56 patients (13 %) without (P = 0.7), and in 14 % of the patients (23/163) with absolute TP vs 1 % with relative TP (1/75), P < 0.01. The results of bone marrow examination are shown in Table 3 and examples of bone marrow smears are shown in Fig. 1.

Table 3 Results of bone marrow aspirates
Full size table
Fig. 1
figure 1

Examples of bone marrow smears (May Grünwald–Giemsa coloration). a, b Hemophagocytosis by medullar macrophages, c megaloblastosis, d increased myeloid precursors

Full size image

Outcome

A total of 44 patients received platelet transfusions during their stay and were therefore excluded from the analysis of risk factors of SBEs. None of them developed SBEs. In the 257 other patients, 38 episodes of SBE were observed in 30 patients. The median nadir platelet count was 44 × 109/L (27–69) in the 30 patients with SBE and 85 × 109/L (56–132) in the 227 patients without (P < 0.0001). SBE was observed in 30 % (26/164) of patients with absolute TP and in 4 % (4/93) with relative (P = 0.01). The incidence of patients developing SBE according to nadir platelet count was 28 % (5/18), 24 % (11/46), 10 % (10/104), and 4 % (4/89) (P = 0.001) for counts of less than 20 × 109/L, between 20 and 49 × 109/L, between 50 and 99 × 109/L, and at least 100 × 109/L, respectively (Fig. 2). In multivariate analysis a nadir platelet count below 50 × 109/L was an independent risk factor of SBEs (P < 0.05) (Supplemental Files 6a and 6b). Transfusion requirements are summarized in Supplemental File 7. Bleeding was the immediate cause of death in four patients with absolute TP. The ICU mortality of thrombocytopenic patients with a nadir platelet count between 50 and 99 × 109/L, between 20 and 49 × 109/L, and at most 20 × 109/L was 22 % (24/109), 43 % (29/67), and 61 % (22/36) (P = 0.009), respectively; the corresponding values for hospital mortality were 30 % (33/109), 52 % (35/67), and 67 % (24/36), respectively (P = 0.01). According to our definitions, the type of TP, absolute or relative, was not associated with ICU and hospital mortality. High SAPS II score, SBE, septic shock, and hemophagocytic syndrome were associated with hospital mortality in multivariate analysis (Supplemental File 8).

Fig. 2
figure 2

Incidence of bleeding (solid bars), packed red blood cell transfusion (horizontal hatch marks), platelet transfusion (open bars), and fresh frozen plasma transfusion (vertical hatch marks) according to nadir platelet count

Full size image

Discussion

Our study is one of the largest on the epidemiology and outcome of thrombocytopenic patients in the ICU setting [3–8, 11, 12, 16–18]. Another major strength of the study is its design. It was a prospective multicenter study in which there was a systematic search for causes of TP in critically ill patients. It also includes the largest number of bone marrow examinations in ICU thrombocytopenic patients performed so far.

Very few data are currently available on bone marrow aspiration results in ICU patients. Two large prospective studies have been performed but included fewer than 50 bone marrow aspirates in septic patients and focused on hemophagocytosis [34, 35]. In other studies bone marrow aspiration was only performed for unexplained TP [12, 17]. In one retrospective study, aspiration was performed in 18/38 thrombocytopenic patients and revealed a decrease in megakaryocytes in 4, but in only 1 did it yield a new diagnosis [12]. In one prospective study, aspiration was performed in 12/52 patients and identified TP etiology in 10 cases [17]. In our work, bone marrow aspiration was performed even when an evident cause of TP was present. Our study suggests that, in a mixed ICU population, TP is rarely due to a decline in platelet medullar production. It added significant information in 22 % of our patients by yielding a diagnosis not previously established. Since we did not record when bone marrow aspiration was performed to exclude a suspected diagnosis, its diagnostic contribution in thrombocytopenic patients could not be accurately measured. However, our study suggests, firstly, that the presence of DIC and sepsis are not sufficient criteria to not perform bone aspiration, particularly in patients with absolute TP, and, secondly, that the results of aspiration performed to assess TP mechanisms in relative TP have such a low impact on management that the procedure can be dispensed with.

Cytological signs suggesting vitamin B12/folate deficiency were observed on bone marrow slides but were not correlated with actual B12/folate deficiency shown by the results of assays. This discrepancy could be related to deficiencies in vitamin B1, vitamin D, and vitamin E or copper [36], which were not measured in our study. Also, we cannot rule out the possibility that in some patients with actual B12 and folate deficiency prior to supplementation the cytological abnormalities of B12 folate deficiency may have persisted. In contrast, the relative rarity of these morphological abnormalities in patients with actual vitamin B12/folate deficiencies could be related to the introduction of automated immunoassays, which yielded an earlier diagnosis of the deficiencies [37].

In our work, the very high rate of TP due to multiple causes may result from the prospective systematic search for an etiology of TP. In numerous cases we identified multiple causes, of which sepsis was the most frequent, as previously reported [6, 16, 17]. An inverse relationship between the severity of sepsis and platelet count has been reported [38]. TP during sepsis may result from an increased consumption or destruction of platelets due to DIC, hemophagocytosis, and immune mechanisms. Additional support for an association between TP and infection derives from studies demonstrating that the platelets significantly influence the balance between infection and immunity [39]. We found a high prevalence of viral infections. However, it is uncertain whether these infections were responsible for the cases of TP.

The main consequence of TP is the perceived risk of bleeding, which may affect the management of thrombocytopenic patients by dissuading physicians from performing invasive procedures and administering anticoagulant therapy and subsequently lead to platelet transfusions. TP may contribute to bleeding in critically ill patients and ultimately result in death [3]. In our study, physicians administered platelet transfusion in patients with a platelet count below 50 × 109/L and who were developing active bleeding or undergoing an invasive procedure. This strategy is in agreement with recommendations published in 2012 by a French expert group [40]. Of note, of 35 patients with a platelet count below 20 × 109/L, 12 did not receive platelet transfusion or develop any severe bleeding events.

Five studies have assessed the risk for bleeding in ICU patients with TP [3, 5, 6, 14, 18], but only two used a consistent definition of bleeding. In univariate analysis most authors found a higher risk of bleeding in patients with lower platelet count [3, 6, 18] but in multivariate analysis the single variable associated with bleeding was sepsis [18]. In our study, the risk of bleeding was assessed after excluding patients who had received prophylactic platelet transfusions. A nadir platelet count below 50 × 109/L was identified as an independent risk factor for bleeding. We were unable to identify TP etiologies and patient subgroups as independent risk factors, but we cannot exclude the possibility that this negative result may have been due to the small number of patients and of SBEs. Furthermore, our study was limited to thrombocytopenic patients and did not include controls, and therefore, it is likely that the risk of severe bleeding events in the absolute TP group was underestimated.

We found that in ICU thrombocytopenic patients the presence of hemophagocytic syndrome carries a poor prognosis. However, we cannot preclude that we had over- or underestimated the diagnosis of hemophagocytic syndrome because we did not measure the absence of natural killer cell activity or the increase in soluble CD25 levels. In fact, hemophagocytic syndrome rather than a factor of excess mortality could be an additional marker of gravity. In a study of 56 patients with hemophagocytic syndrome including 13 with sepsis, the hospital mortality was 50 % and factors associated with hospital survival were Castleman’s disease and B cell lymphoma [35].

Conclusion

TP is associated with multiple etiologies in critically ill patients and, not surprisingly, sepsis is the leading cause. Bone marrow aspirates may yield significant information on TP mechanisms, and contribute to identifying etiologic diagnosis in some specific cases, especially in patients with absolute TP.