Association of platelet count with all‐cause mortality from acute respiratory distress syndrome: A cohort study

Abstract Background The purpose of this study was to investigate whether platelet count was associated with mortality in acute respiratory distress syndrome (ARDS) patients. Methods We analyzed patients with ARDS from Multi‐parameter Intelligent Monitoring in Intensive Care Database III (MIMIC‐III). Platelet count was measured at the time of intensive care unit (ICU) admission. The cox proportional hazard model and subgroup analysis were used to determine the relationship between the platelet count and mortality of ARDS, as well as the consistency of its association. The primary outcome of this study was 365‐day mortality from the date of ICU admission. Result This study enrolled a total of 395 critically ill patients with ARDS. After adjustment for age, gender and ethnicity, the multivariate cox regression model showed that the hazard ratios (HRs) (95% confidence intervals [CIs]) of platelet count <192 × 109/L and >296 × 109/L were 2.08 (1.43, 3.04) and 1.35 (0.91, 2.01), respectively, compared with the reference (192–296 ×109/L). After adjusting for confounding factors, lower platelet count (<192 × 109/L) was associated with increased mortality (adjusted HR, 1.71; 95% CI 1.06–2.76, p = 0.0284). However, there was no similar trend in the 30‐day (adjusted HR,1.02; 95% CI 0.54–1.94) or 90‐day (adjusted HR, 1.65; 95% CI 0.94–2.89) mortality. In the subgroup analysis, lower platelet count showed significant interactions with specific populations (p interaction = 0.0413), especially in patients with atrial fibrillation. Conclusion Taken together, our analysis showed that platelet count is an independent predictor of mortality in critically ill patients with ARDS.

Platelet count is a set of platelet parameters that can be obtained by analysis of standard complete blood counts. 9 Platelet count is mainly used for differential diagnosis of thrombotic diseases. 10 Recent studies, however, have shown that platelets can be used as an independent prognostic indicator of mortality, and are associated with adverse outcomes in diseases such as sepsis, 11 liver failure, 12 multiple myeloma, 13 and blood system diseases. 14 In addition, low platelet count has been shown to be an independent risk factor for increased mortality in critically ill patients. 15 Moreover, available data have demonstrated that platelets can regulate innate and adaptive immune responses, 16 and can be used as a biomarker of both inflammatory and immune responses. [17][18][19] Similarly, ARDS is also associated with immune and inflammatory responses. 20,21 Data on whether platelet count is associated with mortality of ARDS remains scant. There are no data on the relationship between platelet count and prognosis of patients with ARDS. Besides, it is unknown whether platelet count is a risk factor for ARDS in intensive care unit (ICU) patients.
Here, we investigated the effect of platelet count in critically ill patients with ARDS.

| Population selection criteria
The database contains a total of 58,976 critically ill patients and patients with ARDS age of 18 years or older were eligible for enrollment.
The following symptoms were included in the Berlin criteria 24,25 : acute onset, arterial partial pressure of oxygen (PaO 2 )/fraction of inspired oxygen (FiO 2 ) <300 mmHg, positive end-expiratory pressure ≥5 cm H 2 O, and on the first day of admission, the absence of heart failure and the presence of bilateral infiltrates on chest radiograph. The exclusion criteria of the patients from the current study included the following conditions: (1)

| Data extraction
The current study extracted the data using Structured Query Language (SQL) PostgreSQL (version 9.6). The content of the extracted data was demographic information, laboratory parameters, and clinical parameters. The extracted demographic information included: age, gender, ethnicity, systolic blood pressure, diastolic blood pressure (DBP), mean blood pressure, respiratory rate, heart rate, SPO 2 , and temperature. Sequential organ failure assessment (SOFA) and acute physiology score III (APSIII) were all calculated using physiological measurements and clinical information according to published recommendations and accepted formulae. 26 The baseline data were extracted within 24 h after patient admission to the ICU. The primary outcome of the study was the 365-day mortality from the date of ICU admission, whereas 30-day and 90-day mortality after ICU admission were the secondary outcomes.

| PLT assessment
Venous blood samples were collected from subjects within 24 h of admission to the ICU. PLT is measured by medical instruments and expressed as 10 9 /L.

| Statistical analysis
Baseline characteristics of all patients enrolled in the current study were stratified by platelet count tertiles. Furthermore, the chisquare test was used for categorical variables and analysis of variance or Kruskal-Wallis test for continuous variables to compare the groups. In addition, the continuous variables were expressed as mean ± standard deviation or medians, and categorical data were expressed as number or percentage. The cox proportional hazards model was used to determine the association between platelet count and hospital mortality as well as 365-day all-cause mortality for ARDS and the results were expressed as hazard ratios (HRs) with associated 95% confidence interval (CI).
Three models were run for each endpoint in the present study.
All variables included in the multivariable model were selected according to their associations with the outcomes or a change in effect estimate of more than 10%. In model I, no covariates were adjusted.
In model II, covariates were adjusted for age, gender, and ethnicity, whereas in model III, the covariates were further adjusted for DBP, temperature, respiratory rate, heart rate, CHF, chronic liver disease, stroke, pneumonia, vasoactive drug, systemic inflammatory response syndrome (SIRS), PT, and INR. Furthermore, the stratification analyses were also performed to investigate whether the effect of platelets differed across various subgroups, including AFIB, CHF, CAD, CKD, liver disease, and vasoactive drug. The data were analyzed using the R software version 3.42. All statistical analyses were two-sided and p < 0.05 was interpreted as statistically significant.

| Subject characteristics
A total of 395 patients who met the inclusion criteria were divided into three independent groups according to the level of platelet count. Individual characteristics and hematologic laboratory data of the study are presented in Table 1. The patients with a low platelet count value (Platelet count <192 × 10 9 /L) was more likely to be male and have lower heart rate, bicarbonate, glucose, and WBC than those with a high platelet count value (Platelet >296 × 10 9 /L).
Patients with a low platelet count also had higher creatinine, respiratory rate, chloride, sodium, bilirubin, BUN, APTT, and PT than those with a high platelet count. The INR was significantly higher in the group of high platelet count. Moreover, low platelet count was more frequent in patients with liver disease as compared with the rest of the participants.

| Platelet count levels and long-term mortality
It was found that low platelet count was associated with an increased risk of hospital mortality and 365-day all-cause mortality in patients with ARDS ( Table 2). Results of the multivariate Cox regression analysis for 365-day mortality, HRs (95% CIs) of platelet count <192 × 10 9 /L and >296 × 10 9 /L as compared with the reference level (192-296 × 10 9 /L), were 1.82 (1.25, 2.64) and 1.25 (0.84, 1.84), respectively.

| Platelet count levels and 30-day and 90day mortality
For the secondary outcomes of 30-day and 90-day mortality, the results of multivariate analysis showed that the HR (95% CIs) of low platelet count (<192 × 10 9 /L) was 1

| Subgroup analyses
Subgroup analyses were also performed in the present study to examine the consistency of association between platelet count and risk of 365-day hospital mortality in patients with ARDS. It included CHF, CAD, AFIB, liver disease, CKD, and vasoactive drug in the analyses (

| DISCUSS ION
It was evident that low platelet count is an independent risk factor for 365-day all-cause mortality in patients with ARDS in the ICU. In addition, platelet count showed a significant interaction with AFIB in the 365-day mortality of patients with ARDS in ICU. Furthermore, recent studies have shown that platelet count is a new predictor of multiple adverse outcomes. [27][28][29][30][31] The current study also indicated that lower platelet count was an independent predictor of mortality in patients with ARDS in ICU.
The ARDS is a life-threatening complication with high mortality and is a systemic inflammatory response syndrome. 1,20,32 It is associated with multiple organ disorders and systemic inflammation. 2,20 In several previous studies, some biomarkers have been used to predict the prognosis of ARDS. [33][34][35] However, prognostic indicators of ARDS are limited because a single mechanism cannot predict the outcome of a complex syndrome like ARDS. [35][36][37] Platelet count is a low cost and easy to obtain routine item of laboratory examinations. A lot of research studies have shown that platelet count is a reliable predictor of various diseases and organ dysfunction. 17,38 Some studies have found a link between platelet count and platelet count changes in inflammatory markers. 38,39 Furthermore, platelets can limit the growth of bacteria, affect the recruitment and function of white blood cells, as well as cause a series of changes in cytokines. 40,41 Therefore, systemic inflammatory response and multiple organ failure may help explain the potential link between platelet count and mortality in patients with ARDS in the ICU, but the mechanisms are still not clear. The low platelet count could be caused by thrombocytopenia or DIC and both are common complications among critically ill patients and are independent risk factors affecting prognosis. 42,43,38,44 It has been noted that at the setting of severe disease such as sepsis, low platelet count leads to platelet count aggregation, platelet-leukocyte complex formation, and release of molecules that enhance inflammation as well as cell adhesion. 20,44,45 Increases in pro-inflammatory factors and adhesion molecules have a negative impact on the survival of patients. 45,46 Platelet-mediated thrombosis and increased permeability of the endothelial-capillary barrier both reveal a close link between platelet count and ARDS mortality. 10,47,48 To the best of our knowledge, the current research was the first study to fully determine the relationship between platelet count and mortality in patients with ARDS in the ICU. In addition, mortality was chosen as the main outcome.
However, the present study also had some limitations. First, a single-center retrospective design was used with the influence of selection bias. Second, platelet count was measured in patients only upon admission to the ICU and changes in platelet count during ICU hospitalization were not assessed. Third, the number of patients selected was not large. Finally, this study did not consider interaction or nonlinearity for the relationship between covariates and outcome when constructing the model. Ensemble modeling may be needed to address these questions fully. 49

| CON CLUS ION
In conclusion, it was found that platelet count is a novel and independent predictor of mortality in patients with ARDS and in the ICU.
In addition, lower platelet count showed significant interactions with specific populations and especially in patients with atrial fibrillation.
However, there is need for more extensive prospective studies and long-term follow-up to determine the relationship between platelet count and poor prognosis in patients with cancer.

D I SCLOS U R E
The funders of the project were not involved in study design; in the collection; in the data analysis; or in the writing of the report and publication.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the Haiwen Dai (E-mail: m13958001551@21cn.com).