Our clinical question is “What is the evidence for the prognostic value of MPV in adult patients with acute ischemic stroke with any treatment?

We defined MPV as the average size of platelets in circulation expressed in femtoliters (fL). A normal MPV has a range of 7.5-11.5 fL [24]. Also, we ranked the importance of the outcomes, according to the literature, as follows [6]:

1) All-cause mortality rate

2) Functional outcome (We used a dichotomic approach with good functional outcome assessed with modified Rankin scale [mRS] ≦2) and poor functional outcome (mRS >2)

3) Impairment measure (evaluated by the National Institutes of Health Stroke Scale (NIHSS)

4) Complications associated with stroke during hospitalization

We searched in PubMed, Embase, Scopus, Web of Science, and Google Scholar until March 2024. The search strategy was based on controlled (MeSH and Emtree) and free terms (Supplementary Material S1). There were no restrictions on language or publication date.

Studies were included if they met the following criteria: 1) adult participants (aged >18 years old); 2) Studies that evaluated the association between MPV and any motor or non-motor outcome; 3) Analytical observational studies (cross-sectional, case-control and cohort studies). We excluded narrative reviews, studies in non-humans, case reports, conference abstracts, letters, and editorials.

The electronic search results were imported into Endnote X9, and duplicate records were excluded. Then, these records were exported to the Rayyan (https://rayyan.qcri.org/). A peer review process was performed by two reviewers (MCT and RVE), and any discrepancies were resolved by consensus and with the opinion of the third diriment (CAD or JS). These reviewers assessed inclusion criteria independently by reading the full texts of the potentially relevant studies selected, and discrepancies were resolved according to consensus. The complete list of excluded articles is provided in Supplementary Material S2.

Two groups of authors (MPG and RVE; MCT and CQV) independently carried out data extraction using a form, and any disagreements were resolved by consensus or by a third author (CAD or JS). We extracted the following information: title of the study, first author, year of publication, study design, country where the study was performed, number of participants, sex, age, sample time, mean or median MPV according to sample stratification, follow-up time, crude and adjusted association measures, and outcomes. If additional data were needed, we contacted the corresponding author through email to request further information.

The quality of the studies was assessed with the Newcastle Ottawa Scale (NOS) [31] by two authors (MCL and CCB). This tool evaluates the quality of published nonrandomized studies and is based on three items: selection, comparability, and outcome/exposure. Each item has sub-items, on which a star-based score was assigned. Studies with scores ≥6 were considered as having a low risk of bias (high quality), scores of 4-5 as having a moderate risk of bias, and scores <4 as having a high risk of bias.

A meta-analysis was planned for each outcome; however, when this was not possible due to unavailable data, narrative synthesis was performed. The standard deviation of the difference in means was calculated with the method reported in the Cochrane Handbook for reporting narrative synthesis [13]. When authors reported with median and interquartile range, Wan’s method was used for conversion to mean and standard deviation [29].

Meta-analyses were performed using the inverse variance method and the randomized effects mode and the variance between studies (τ2) was estimated using the DerSimonian-Laird estimator [20]. Mean differences (MD) with their 95% confidence intervals (CI 95%) were pooled. Heterogeneity between studies was assessed using the I² statistic. Heterogeneity was defined as low if I² <30%, moderate if I²=30-60%, and high if I² >60%. The metacont function of the meta-package in R 4.1.0 was used (www.r-project.org). We performed subgroup analyses by reported therapy. Finally, we performed a leave-one-out sensitivity analysis, and others, using the Paule-Mandel estimator. This decision was made taking into consideration that the DerSimonian-Laird estimator underestimates the true heterogeneity when τ2 is large [30].

To interpret the correlation coefficients, we adopted the framework presented by Schober et al. where a coefficient of 0.00-0.09 was considered negligible, 0.10-0.39 as weak, 0.40-0.69 as moderate, and 0.70-0.89 as strong correlation.

Two authors (CCB and CQV) assessed independently the certainty of our pooled results and qualitative synthesis applying the Grading of Recommendation, Assessment, Development, and Evaluation (GRADE) on continuous outcomes, narrative synthesis [21], and prognostic factors [14]. This assessment is based on five domains: study limitations (risk of bias of the studies included), imprecision (sample size and confidence interval), indirectness (generalizability), inconsistency (heterogeneity), publication bias, large magnitude of an effect, dose-response gradient, and effect of plausible residual confounding as stated in the GRADE handbook [25]. We adapted the GRADE criteria to our results. The certainty of the evidence was characterized as high, moderate, low, or very low. For assessing imprecision criteria in coefficient correlation, we interpret negligible and weak correlation as imprecise. Publication bias was assessed through funnel plots, Egger’s and Begg’s tests.

We identified 1,048 studies through our systematic search. We removed duplicates and screened 473. Finally, we included 57 articles (Fig. 1).

Twelve studies were prospective cohorts, forty-four studies were retrospective cohorts, and one was a case-control study. The total number of participants was 10979. The average age range of patients with ischemic stroke was between 49.9 and 77.59. The summary of the characteristics of studies is summarized in Supplementary Material S3.

Eleven studies, which included 1978 participants, reported mortality with a follow-up time between 7 days and 1 year. The MD of MPV between the patients who died and survived varied between -0.26 and 7.06. Only one study reported outcomes after 3 months. Ghodsi et al reported mortality at one year of follow-up with an OR: 1.76 (95% CI: 1.43 to 2.09) and a cut-off point of MPV: 9.54 (Table 1).

Among the 57 studies that met our inclusion criteria, fourteen were pooled to assess mRS in 90-day follow-up. A total of 3,563 participants were selected, of which 1,353 had a poor functional outcome (mRS >2). In the pooled analysis, a significant increase in MPV values was observed with high heterogeneity (MD: 0.50; 95% CI: 0.31 to 0.70; I²=82%) (Fig. 2).

According to the subgroup analysis. A significant increase in MPV values was observed in studies with patients with intravenous thrombolysis (MD: 0.37; 95% CI: 0.13, 0.60; I²=74%) and in those that did not determine their treatment (MD: 0.70; 95% CI: 0.33, 1.06; I²=89%) (Fig. 2).

Regarding our sensitive analysis, when single studies were sequentially removed, no significant variation in the pooled MD was observed (between 0.43 and 0.54) (Supplementary Material S4). Also, no significant variation was observed using the Paule-Mandel estimator (MD: 0.51; 95% CI: 0.28; 0.75; I²=82%) (Supplementary Material S5). This suggested that our estimation was stable.

Among the 39 studies that met our inclusion criteria, three were pooled to assess mRS in a one-year follow-up. A total of 1090 participants were selected, of which 262 had poor functional outcomes (mRS >2). In the pooled analysis, a non-significant increase in MPV values was observed with high heterogeneity (MD: 0.70; 95% CI: -0.19 to 1.59; I²=94%) (Fig. 3).

Regarding our sensitive analysis, when single studies were sequentially removed, no significant variation in the pooled MD was observed (between 0.23 and 1.03) (Supplementary Material S6). Also, no significant variation was observed using the Paule-Mandel estimator (MD: 0.70; 95% CI: -0.15 to 1.55; I²=94%) (Supplementary Material S7). This suggested that our estimation was stable.

Three studies reported NIHSS as an outcome with follow-up between 24 hours and 3 months after symptom onset. The range of correlation coefficients was weak (r=0.140-0.221) (Table 2).

We performed the risk of bias in the 57 included studies. Two and eight articles had “high risk of bias” and “moderate risk of bias”, respectively. D’Erasmo et al obtained the lowest score: 2/9 because the cohort was not representative, there was no definition of the diagnostic criteria, nor did they clearly report the statistical analysis (Supplementary Material S10).

Begg’s test (z=2.90, p=0.004) and Egger’s test (t=2.70, p=0.020) revealed the potential publication bias of the included literature, and the funnel plot showed an asymmetry in the Activity measure at three months (Supplementary Material S11).

We started with high certainty because all the studies included were comparative observational. We considered mortality at one year to be high with certainty. With moderate evidence, we find mortality up to three months, NIHSS, cerebral microbleeds, and complications during hospitalizations due to the imprecision of the results. Measurement of activity at one year and hemorrhagic transformation were evaluated with low certainty due to the I² being greater than 60% and the imprecision. With very low certainty of the evidence, the measurement of activity at 3 months was evaluated due to inconsistency, imprecision, and risk of publication bias (Table 3).

This systematic review with meta-analysis (10,979 participants included) revealed that high MPV levels in AIS patients were associated with higher mortality at 90 days and 1 year. Furthermore, we found a significant increase in MPV in AIS patients with worse functional outcome at 90 days. Additionally, with moderate certainty, our results showed a weak correlation between high MPV levels and impairment severity. Lastly, we found no significant difference in MPV levels for AIS patients with worse functional outcomes at one year, with low certainty, as well as complications like cerebral microbleeds, hemorrhagic transformation, and complications during hospitalization.

Our pooled analysis revealed that MPV was increased in the poor functional outcome group when compared with the favorable functional outcome; this finding is in line with the results of other systematic reviews of AIS patients. Zheng et al. [33] reported in a meta-analysis of 10 articles (2,390 patients) that lower MPV values [Standardized Mean Difference (95% CI)=-0.52 (-0.80, -0.24)] were present in a favorable functional outcome. This may be explained due to the fact that continuous production and activity of platelets has been described just up to 6 months after the stroke [2]. On the other hand, we found that MPV levels were associated with short- and long-term mortality. This finding contrasts with a systematic review and meta-analysis in critically ill patients, where no significant correlation was found between initial MPV and in-hospital death [27]. This could be explained because most of the studies included in this review included patients with sepsis, which could cause platelets to be depleted in severe and late forms.

Regarding adverse outcomes such as impairment measure, hemorrhagic transformation, or cerebral microbleeds, we did not find significant differences in MPV between the groups. However, other reviews reported different biomarkers (MMP-9, neutrophil to lymphocyte ratio, and natriuretic peptide) associated with hemorrhagic transformation and heart failure [17]. This difference could be because these biomarkers are associated with high hyperacute inflammation. However, it has been described that in the phase of acute inflammation, the MPV increases and then decreases when larger platelets are used in the inflammatory process. Additionally, it has been described that MPV levels are not directly related to the increase in white blood cell count in acute inflammation [1].

According to the GRADE approach, the certainty of the evidence for MPV as a prognostic factor for most of the outcomes ranged from very low to moderate due to imprecision of the results, heterogeneity, inconsistency, and risk of publication bias. This could have been because most of the studies were retrospective, so their results could be prone to bias. Additionally, most of the included studies did not report the type of treatment, which could affect the patient’s prognosis, and four of our results rely on a single study, meaning the certainty of the evidence may vary as more research is published. Future prospective studies are essential to address these limitations, ensuring better control of confounding variables, appropriate follow-up durations, and adequate sample sizes to establish a clear MPV cut-off point and enhance result precision. Finally, we advocate for the development of prognostic models that include readily available biochemical variables, such as MPV, which could help predict adverse outcomes in these patients.

The mean platelet volume (MPV) is a laboratory marker associated with platelet function and activity. Increased MPV in thromboembolic disease is reflected as an important risk factor [7]. This biomarker has emerged as a promising biomarker in neurology, reflecting inflammatory processes and vascular dysfunction associated with various neurological disorders. The value that this biomarker can have in differentiating patients with stroke and healthy controls has already been described [23]. Its effectiveness lies in its ability to provide valuable insights into various neurological conditions at a relatively low expense. Our results indicate that MPV could serve to predict mortality and functional outcome. In that sense, we recommend taking MPV into account in clinical practice to elucidate the medium and long-term prognosis of AIS patients. As a routine part of a complete blood count test, its affordability, coupled with its diagnostic and prognostic significance, makes MPV a useful tool in healthcare. However, it must be considered that outcomes during hospitalization were not found to be associated with MPV levels. Therefore, MPV interpretation should be done in conjunction with other clinical and laboratory findings for accurate diagnosis and management.

Our systematic review has some limitations. First, most of the included studies were from Asia, which could affect the generalizability of our results. In addition, Chinese databases were not included in the search strategy. Furthermore, our meta-analysis was based on unadjusted mean differences, so the results may be biased. However, our results were strengthened by sensitivity analyses performed. Despite these, our study also has strengths. We conducted a comprehensive systematic search without language or time restrictions. In addition, this is the systematic review evaluating the prognostic value of MPV in AIS patients with the largest number of participants taking part in the meta-analysis and including participants from over 15 countries. Additionally, we evaluated both short- and long-term outcomes compared to previous systematic reviews. Finally, we evaluated the certainty of our results using the GRADE criteria.