The day after intracerebral hemorrhage: platelet mass index as predictor of survival—a retrospective cohort study

Abstract

Background

Platelets are implicated in the pathophysiology of intracerebral hemorrhage (ICH). Platelet count (PLT) is affected by platelet loss, while mean platelet volume (MPV) by platelet replenishment. Whether platelet mass index (PMI), the product of PLT and MPV, might predict survival after ICH, remains unknown.

Methods

All first-ever ICH patients, admitted to Xanthi General Hospital between January 2018 and May 2020 and met eligibility criteria, were enrolled in this retrospective cohort study. Demographics, medical record, first-symptom-to-admission time, vital signs, modified Rankin Scale, ICH score, arterial blood gas test, complete blood count, blood biochemistry, and CT scan test were collected for each patient. PMI values on day 1 (admission; PMI₁), day 2 (PMI₂), and day 7 (PMI₇), along with PLT, MPV, platelet distribution width (PDW), and platelet large cell ratio (P-LCR), were evaluated as potential predictors of 12-month survival using Repeated Measures General Linear Model. Binary discretization of predictors was based on optimal scaling and evaluated using binary regression.

Results

From 59 patients enrolled (aged 75.7 ± 12.0 years; 31 females), 29 were still alive 12 months after ICH. Age, arterial hypertension, diabetes mellitus, hemoglobin level (Hb), and oxygen saturation (O₂Sat) were correlated with 12-month survival. After adjustment for these parameters, PMI₁ and PMI₂ were independently correlated with 12-month survival (P = 0.048 and P = 0.004, respectively), while PMI₇ was not (P = 0.332). PMI₂ ≥ 2,400 fL/μL was best to discriminate survivors from non-survivors (age, arterial hypertension, diabetes mellitus, Hb, and O₂Sat adjusted OR 0.123 with 95% CI: 0.023–0.694; P = 0.018).

Conclusions

PMI within the first day after admission for ICH might be used as early predictors of survival. Properly designed prospective studies are needed to further evaluate their contribution as such.

Background

Intracerebral hemorrhage (ICH) is a major public health issue causing high rate of mortality as well as disability [1]. ICH is the second commonest subtype of stroke, accounting for 10–20% of all stroke cases [2]. Increased age, male sex, hypertension, diabetes, and high alcohol intake are well-described risk factors for ICH [3, 4]. Anticoagulant and antiplatelet treatment also increase the risk for ICH [5, 6], and platelet count (PLT) might be a true risk factor for anticoagulant-associated ICH [6].

Many scoring systems have been developed for the prediction of outcome in ICH. The ICH score uses Glasgow Coma Scale (GCS), age ≥ 80 years, infratentorial origin of ICH, ICH volume, and presence of intraventricular hemorrhage to assess 30-day mortality [7]. The FUNC score incorporates age, GCS, ICH location, ICH volume, and pre-ICH cognitive impairment το predict functional outcome 3 months after ICH [8]. Moreover, single parameters, as low hemoglobin (Hb) level [9], hyperglycemia [10], and increased serum C-reactive protein [11] have also been recognized as potential prognosticators of poor outcome or death in ICH.

It has been demonstrated that PLT was significantly lower in hemorrhagic strokes when compared with controls [12]. PLT on the first day after admission, when considered as scale variable, might be a good predictor of mortality in hemorrhagic stroke [13]. Moreover, PLT has been proposed as an independent predictor of poor outcome at time of discharge in cerebellar hemorrhage [14]. Of note, thrombocytopenia (PLT < 150•10^–9/L) itself did not affect functional outcome among ICH independently of antiplatelet treatment [15]. As far as platelet indices are concerned, only scarce evidence regarding their potential role of in ICH is available. Increased (> 13 fL) mean platelet volume (MPV) has been associated with hemorrhagic stroke, when compared to controls; nevertheless, neither MPV nor PLT were associated with outcome prognosis [16]. Platelet mass index (PMI), which is the product of PLT and MPV, has been proposed as predictor of intraventricular hemorrhage in very-low birth-weight newborns [17].

Despite the emerging evidence that platelets are contributing to the risk for ICH, their role remains partly understood. On the other hand, it is known that alterations of PLT and platelet indices including MPV, platelet distribution width (PDW), and platelet large cell ratio (P-LCR; the percentage of platelets with volume > 12 fL) are observed in platelet activation, consumption and replenishment [17].

The present study aimed to investigate whether initial measurements as well as dynamic changes in PMI can be of predictive value regarding 12-month survival after ICH.

Methods

Study design

This retrospective cohort study was conducted at Xanthi General Hospital between January 2018 and May 2020. The study protocol was approved by the Scientific Board of Xanthi General Hospital (Decision No. 103/May 17, 2021). The report was prepared according to the STROBE guidelines [18].

Patients

All consecutive patients, admitted for ICH at Xanthi General Hospital between January 2018 and May 2020, were considered for eligibility. Inclusion criteria were: a) age ≥ 18 years, b) first-ever episode of ICH. Exclusion criteria were: a) history of previous stroke, b) treatment with anti-rheumatic agents, and c) hospitalization for less than 24 h after admission. The follow-up period for each patient was set to 12 months after the episode of ICH. Follow-up was aided by the use of public insurance records. Demographics, medical record, first-symptom-to-admission time, vital signs, modified Rankin Scale (mRS), ICH score, arterial blood gas test, complete blood count, blood biochemistry, and computed tomography (CT) scan test were collected for each patient. PMI values on day 1 (admission; PMI₁), day 2 (PMI₂), and day 7 (PMI₇), along with PLT, MPV, PDW, and P-LCR values, were evaluated as potential predictors of 12-month survival.

CT scan

The CT scan was used to assess the presence of intraventricular hemorrhage, determine the origin of ICH (Infratentorial or not), and compute the ICH hematoma volume in cm³ by two separate specialists. In detail, the ½*(A)*(B)*(C) formula was preferred, where A represents the greatest diameter on the largest hemorrhage slice in cm, B the diameter perpendicular to A in cm, and C the approximate number of axial slices with hemorrhage multiplied by 0.5 cm, namely the slice thickness [7].

Statistical analysis

To compare scale as well as nominal variables between different outcomes, Student’s t-test and Chi-square test were used respectively. However, regarding the latter, the Fisher exact test was alternatively used in cases that expected frequencies were ≤ 5 in ≥ 25% of cells. Correlations were assessed using the Pearson’s parametric correlation coefficient.

The repeated measures General Linear Model was used for analysis of within-subject and between-subject variance of the same variable measured more than once in each patient.

Optimal cut-offs were determined by transformation of scale variables to binary ones through optimal scaling; in detail, discretization to two groups, regularization using ridge regression, and tenfold cross-validation were performed through SPSS CATREG procedure.

To explore the potential value of PMI₂ to predict outcome as binary variable independently of age, diabetes mellitus, hypertension, Hb, and sPO₂, multivariate analysis was performed using binary regression model (the probability for stepwise entry and removal were set to 0.05 and 0.10, respectively; the classification cutoff was set to 0.5; the maximum number of iterations was set to 20).

Sample sizes were selected to detect less than 20% difference tolerating 0.05 type I error and 0.10 type II error. For that purpose, the relevant on-line tool freely available at https://www.stat.ubc.ca/~rollin/stats/ssize/n1.html. was used.

Descriptive statistics are provided either as means along with their relevant standard deviations, or percentages, for scale and nominal variables respectively. All reported p values are two-sided. The level of statistical significance was set to p = 0.05. All numerical values are given with at least two significant digits. Missing data were excluded. Statistical analysis and visualization of results was performed with the use of IBM SPSS Statistics software, version 26.0, for Windows; MedCalc Version 20.218 (MedCalc Software Ltd; 2023) was used to illustrate forest plots.

Results

Fifty-nine patients (31 women), aged 75.7 ± 12.0 years, were enrolled. Twenty-nine were still alive 12 months after ICH. As far as non-survivors are concerned, 28/30 (93.3%) succumbed within the first 30 days after ICH. In detail, the leading causes of death were arrhythmia (11/30; 36.7%), infection/sepsis (8/30; 26.7%), and cardiovascular disease (7/30; 23.3%), followed by respiratory failure (2/30; 6.7%) and recurrent hemorrhage (2/30; 6.7%). A flow diagram is provided as Fig. 1.

Fig. 1

Flow diagram of the study

Characteristics of patients and detailed comparisons between 12-month survivors and non-survivors are presented in Table 1. Of note, 4/59 patients (6.8%) presented thrombocytopenia due to infection/sepsis (1 case), liver cirrhosis (1 case), myelodysplastic syndrome (1 case), and bone marrow infiltration from metastatic lung cancer (1 case)". No patient received platelets, fresh frozen plasma, or any other treatment for thrombocytopenia.

Table 1 Patients’ characteristics and univariate analysis based on 12-month survival status

Univariate analysis demonstrated that younger age (P = 0.001), absence of diabetes mellitus (P = 0.013), absence of arterial hypertension (P = 0.019), elevated Hb (P = 0.019), elevated glucose (P = 0.021), and increased sPO₂ at admission (P = 0.023) were correlated with survival (Table 1). Moreover, increased PMI₂ (P = 0.020) and PLT at day 2 (P = 0.017) were correlated with survival (Table 2).

Table 2 Platelet number and indices (measured at admission, day 2, and day 7), as well as univariate and multivariate analysis based on 12-month survival status

Aiming to elucidate whether PMI values might be used as early predictors of survival, repeated measures General Linear Model was used. In that model, PMI₁ and PMI₂ were independently correlated with survival (P = 0.048 and P = 0.004, respectively), while PMI₇ was not (P = 0.332), after adjustment for age, diabetes mellitus, arterial hypertension, Hb, and sPO₂ at admission (Table 2, Fig. 2). Glucose was initially excluded from adjustment due to collinearity issues attributed to diabetes mellitus.

Fig. 2

Estimated Marginal Means of 12-month survival at mean of covariates (age, hypertension, diabetes, Hb, and sPO₂) for PMI₁ (day 1; admission), PMI₂ (day 2), and PMI₇ (day 7) using Repeated Measures General Linear Model

To further investigate the contribution of additional potential confounders to within-samples variability concerning the consecutive measurements of PLT and PMI, Repeated Measures GLM multivariate models based on 12-month survival status adjusted for age, diabetes, hypertension, Hb, and sPO₂ were performed (Table 3). These models suggested that hyperlipidemia, glucose, lactates, and temperature, but not gender, history of coronary artery disease (CAD), antiplatelets, and anticoagulants, might constitute true confounders. PMI₂ were still independently correlated with survival (P = 0.012) after adjustment for these additional confounders (Table 2).

Table 3 PLT and PMI (measured at admission, day 2, and day 7) multivariate analysis based on 12-month survival status additionally adjusted for age, diabetes, hypertension, Hb, and sPO₂ using Repeated Measures GLM: Contribution of potential confounders to within-samples variability

Binary discretization of PMI₂, after adjustment for age, diabetes mellitus, arterial hypertension, Hb, and sPO₂ at admission, suggested 2,400 fL/μL as cut-off (Fig. 3). Using binary regression, PMI₂ ≥ 2,400 fL/μL was independently correlated with survival (OR 0.123; 95% CI: 0.022–0.694; P = 0.018), after adjustment for age (OR 1.872 per decade; 95% CI: 0.959–3.655; P = 0.066), diabetes mellitus (OR 3.527; 95% CI: 0.231–53.910; P = 0.365), arterial hypertension (OR 9.837; 95% CI: 1.318–73.426; P = 0.026), Hb (OR 0.395 per g/dL; 95% CI: 0.187–0.834; P = 0.015), and sPO₂ at admission (OR 0.664 per %; 95% CI: 0.434–1.016; P = 0.059) (Fig. 4).

Fig. 3

Ridge paths determining the best cut-off for PMI₂ (2,400 fL/μL)

Fig. 4

Forest plot depicting binary regression model for PLT₂ ≥ 2,400 fL/μL; OR < 1 favors 12-month survival

PLT values, when measured at day 2, were also independently correlated with survival (P = 0.017); however, PLT values, when measured at admission and day 7, failed to demonstrate prognostic value (P = 0.069 and P = 0.460, respectively). Moreover, MPV, PDW, and P-LCR had no prognostic value (Table 2).

Of note, PLT₂ (r = − 0.312; P = 0.022), and PMI₂ (r = − 0.285; P = 0.038) were negatively correlated with mRS. Moreover, these parameters presented a weaker, yet significant negative correlation with ICH score (PLT₂: r = − 0.278; P = 0.042), and PMI₂: r = − 0.275; P = 0.046) (Fig. 5A–D).

Fig. 5

Scatterplots depicting correlations between A PLT₂ with MRs (left column; upper row), B PMI₂ with MRs (right column; upper row), C PLT₂ with ICH score (left column; lower row), and D PMI₂ with ICH score (right column; lower row)

Discussion

Our study has demonstrated that increased PMI₁, PMI₂, and PLT₂ values were correlated with 12-month survival after first-ever episode of ICH independently of age, diabetes mellitus, arterial hypertension, hemoglobin level, and sPO₂ at admission. PMI proved to be earlier and stronger prognosticator. The most accurate binary predictive model was based on PMI₂, using ≥ 2,360 fL/mL as cut-off. To the best of our knowledge, this is the first time that PMI has been studied, in parallel with PLT and platelet indices, at three different time-points during hospitalization (admission, day 2, and day 7), as potential predictors of 12-month survival after ICH. In detail, increased PMI at both admission and day 2 characterize survivors.

Interestingly, the use of both PLT₂ and PMI₂ as prognosticators of outcome is in keeping with well-established and widely used clinical scores, namely the mRS and the ICH score.

Lower PLT is a known risk factor for ICH, while even lower PLT₂ values in non-survivors reflect platelet consumption [12, 13]. Moreover, it has been proposed that platelet consumption and hyperreactivity coexist in animal models of experimental traumatic hemorrhage [19]. ICH, as a clinical analogue of these models in humans, is accompanied by platelet hyperdestruction, followed by immediate onset of reactive megacaryopoiesis, and production of young, large, and reactive platelets; the younger the replenished platelets, the bigger and more reactive they are [20]. This is in keeping with increased PMI values, a phenomenon that is more pronounced at day 2 and wanes at day 7. In fact, PMI has been introduced as a composite marker to simultaneously assess platelet destruction (attributed to PLT) and replenishment (attributed to MPV). The present study proposed that the larger their value, the better prognosis an ICH patient has for 12-month surviving.

There is evidence suggesting that increased platelet number and reactivity, as reflected mainly by increased MPV and PLT, and thus PMI, shares a strong genetic component influencing variation in platelet reaction at the site of vessel wall injury [21]. Platelet reactivity might be further perplexed by ICH-induced post-traumatic vasospasm, occurring mainly in cases of intraventricular hemorrhage [22]. Moreover, emerging megakaryopoiesis and thrombopoiesis during the ICH acute phase may alter the bone marrow environment affecting the molecular signature of platelets and other cellular compartments [23, 24]. The effect of these alterations on injured brain tissue, locally and/or systematically, is largely unknown and thus remains to be elucidated.

The well-defined limitations of a retrospective cohort study, such as the absence of data on potential confounding factors and the difficulty to identify study and control groups, remain a concern; the present study lacks credible data concerning smoking habits and alcohol consumption, though not decisive for the quality of the analysis. On the other hand, PLT and platelet indices are immediate and inexpensive parameters that are widely available in almost any health setting.

Conclusions

As a conclusion, PMI values at admission and second day of hospitalization might be used as early predictors of survival in ICH, rendering the measurement of PLT, and MPV a valuable and inexpensive tool. Properly designed prospective studies are needed to further evaluate their contribution as such.