Association between circulating inflammatory biomarkers and functional outcome or perihaematomal oedema after ICH: a systematic review & meta-analysis

Background Currently, there are no specific medical treatments for intracerebral haemorrhage (ICH), but the inflammatory response may provide a potential route to treatment. Given the known effects of acute brain injury on peripheral immunity, we hypothesised that inflammatory biomarkers in peripheral blood may be associated with clinical outcome following ICH, as well as perihaematomal oedema (PHO), which is an imaging marker of the neuroinflammatory response. Methods We searched OVID Medline and EMBASE on 07 April 2021 for studies of humans with ICH measuring an inflammatory biomarker in peripheral blood and PHO or clinical outcome. Risk of bias was assessed both by using a scale comprising features of the Newcastle-Ottawa Assessment Scale, STROBE-ME and REMARK guidelines, and for studies included in meta-analysis, also by the QUIPS tool.We used random effects meta-analysis to pool standardised mean differences (SMD) if ≥1 study quantified the association between identical biomarkers and measures of PHO or functional outcome. Results Of 8,615 publications, 16 examined associations between 21 inflammatory biomarkers and PHO (n=1,299 participants), and 93 studies examined associations between ≥1 biomarker and clinical outcome (n=17,702 participants). Overall, 20 studies of nine biomarkers (n=3,199) met criteria for meta-analysis of associations between inflammatory biomarkers and clinical outcome. Death or dependency (modified Rankin Scale (mRS) 3‒6) 90 days after ICH was associated with higher levels of fibrinogen (SMD 0.32; 95%CI [0.04, 0.61]; p=0.025), and high mobility group box protein 1 (HMGB1) (SMD 1.67; 95%CI [0.05, 3.30]; p=0.04). Higher WBC was associated with death or dependency at 90 days (pooled SMD 0.27; 95% CI [0.11, 0.44]; p=0.001; but the association was no longer significant when the analysis was restricted to studies with a low risk of bias (pooled SMD 0.22; 95% CI -0.04-0.48). Higher CRP seemed to be associated with death or dependency at 90 days (pooled SMD 0.80; 95% CI [0.44, 1.17]; p<0.0001) but this association was no longer significant when adjusted OR were pooled (OR 0.99 (95% CI 0.98-1.01)). Conclusions Higher circulating levels of, fibrinogen and HMGB1 are associated with poorer outcomes after ICH. This study highlights the clinical importance of the inflammatory response to ICH and identifies additional research needs in determining if these associations are mediated via PHO and are potential therapeutic targets. Registration PROSPERO ( CRD42019132628; 28/05/2019).

(SMD) if ≥1 study quantified the association between identical biomarkers and measures of PHO or functional outcome.

Conclusions
Higher circulating levels of, fibrinogen and HMGB1 are associated with poorer outcomes after ICH.This study highlights the clinical importance of the inflammatory response to ICH and identifies additional research needs in determining if these associations are mediated via PHO and are potential therapeutic targets.

Introduction
Spontaneous intracerebral haemorrhage (ICH) accounts for 16-30% of all stroke cases 1 .Overall, 40% of patients die within one month after ICH with only 12-39% living independently after six months 2 .There are currently no specific medical treatments that have proven benefit to improve outcome 3 .
There is growing interest in the inflammatory reaction after ICH since it may be a therapeutic target 4 .Perihaematomal oedema (PHO) is thought to be an imaging biomarker of the neuroinflammatory response 5 and neuroinflammation may lead to PHO development.PHO is thought evolve in stages with initial cytotoxic oedema leading to vasogenic oedema 5 .Inflammatory biomarkers are highly expressed in human brain tissue following ICH 6 and several ongoing clinical trials are targeting the immune system 7 .In animal models, local and recruited immune cells release inflammatory mediators (e.g., cytokines, matrix-metalloproteinases (MMPs), and damage associated molecular patterns (DAMPs)) that contribute to brain damage and repair 8 .Various immunomodulatory treatments have been shown to improve outcome in preclinical models of ICH 9 .In humans, stroke-induced changes to the peripheral immune system are associated with the development of sequelae.Circulating immune cell characteristics are associated with trajectory of post-stroke cognitive impairment 10 and fewer circulating leukocytes are associated with infection following ICH 11 .Since PHO is an imaging marker of neuroinflammatory response, we hypothesised that circulating inflammatory biomarkers associated with clinical outcome should also be associated with measures of PHO.
We therefore conducted a systematic review and meta-analysis of associations between circulating inflammatory biomarkers and either PHO or functional outcome after ICH.We believe this is the first systematic review and meta-analysis that has assessed the relationship between inflammatory biomarkers and PHO and/or functional outcome after ICH without restriction based on pre-specified biomarkers.We aimed to (i) identify all studies examining associations between one or more blood-based biomarker and PHO or functional outcome after ICH, (ii) quantify pooled associations between individual biomarkers and PHO or outcome where possible, and (iii) determine whether study-level variables modified any of these associations.

Methods
We performed a systematic review and meta-analysis.The study protocol was pre-registered with the International prospective register of systematic reviews (PROSPERO; CRD42019132628; 28 May 2019).This systematic review is reported in line with the PRISMA guidelines 12 .
Search strategy and study selection CK searched Ovid MEDLINE (1946) (RRID:SCR_002185) and Ovid EMBASE (1974) (RRID:SCR_001650) on 07 April 2021 using a prespecified search strategy that consisted of terms to identify studies that measured inflammatory markers in the blood of ICH patients and assessed functional outcome and/or PHO (Supplemental Information 1, available as Extended data 13 ).The search was not limited by language or publication date.After automated de-duplication in Covidence (covidence.org),three authors (CK, LS, JB) independently screened titles and abstracts to identify potentially eligible studies and read the full text of articles that were potentially eligible for inclusion.Corresponding authors were contacted if full-text articles could not be obtained.A third independent reviewer (FHBMS, JL or NS) made the final decision over inclusion when conflicts arose at the abstract or full-text screening stage (Figure 1).

Eligibility criteria
We included observational studies of ≥5 adults (≥16 years of age) with spontaneous ICH where inflammatory markers were measured in the blood, serum or plasma.We selected biomarkers if they were a marker (e.g., C-reactive protein) or a mediator (e.g., cytokines) of the inflammatory response and grouped the inflammatory biomarkers into six categories based on broad biological activity 14 .Categories were defined based on discussions between domain experts (JB, BM): immune cells, acute phase reactants (defined as liver-produced plasma proteins 15 ), cytokines/chemokines, damage-associated molecular patterns (DAMPs-defined here as molecules released from intracellular compartments during inflammation 16 ), tissue remodelling factors and adhesion molecules.We excluded: 1) studies of ICH due to an underlying macrovascular cause, traumatic ICH, ICH due to hereditary cerebral amyloid angiopathy

Amendments from Version 1
Thank you to the reviewers for their thoughtful and insightful reviews of our article.
The most significant change to our article is in relation to the risk of bias assessment of articles included in the meta-analysis of circulating inflammatory biomarkers and functional outcome after ICH.We have now included a risk of bias assessment of these 20 articles using the QUIPs tool which was specifically designed to assess the risk of bias in prognostic factor research.This can be viewed in Supplemental Table 1 in the Extended Data (https://doi.org/10.6084/m9.figshare.24559081.v1).In the original submission if more than 10 studies were included in the meta-analysis, studies were grouped by study quality based on our original risk of bias score.These analyses have been re-run having stratified the studies into low/moderate/high risk of bias using their QUIPs rating.Importantly, the association between white blood cells (WBCs) and clinical outcome was no longer statistically significant in studies with a low risk of bias.
We have also now included information on whether studies included an adjusted or unadjusted odds ratio and where at least two studies provided these, a meta-analysis of these studies were included, although this was often not possible.Of note, higher C-reactive protein seemed to be associated with worse clinical outcomes when standardised mean differences were pooled, but this association was no longer significant when adjusted odds ratios were pooled.Additionally, a comment on the number of studies that included a multivariable analysis has been added to the results section for each of the biomarkers that were metaanalysed, highlighting the need for more multivariable analyses in this area of research.

Any further responses from the reviewers can be found at the end of the article
or studies of ICH with mixed causes, 2) stroke cohorts where spontaneous ICH cases could not be separated from non-ICH cases, 3) studies including surgically treated patients that could not be separated from non-surgically treated cases, 4) conference abstracts, systematic or narrative reviews.Where studies had overlapping cohorts, the study with the largest cohort was included.

Data extraction
A minimum dataset (the biomarker, outcome definition, time point of biomarker and outcome measurement, measure of association between biomarker and outcome and risk of bias) was extracted from all included studies by one reviewer (CK, JB or NS) and is summarized in a narrative synthesis (Supplemental Table 2, Extended data 13 ).Two authors (CK, JB, or NS) used a standardized proforma to independently extract data from all studies included in meta-analysis (Supplemental Information 2 13 ).Conflicts were arbitrated by a third reviewer (JB, NS).As the included studies were mostly case series with no control group, risk of bias was assessed by one reviewer (CK, JB, NS) using an eight-item composite scale comprising features of the Newcastle-Ottawa Assessment Scale 17 (case definition, representativeness of the cases), STROBE-ME 18 and REMARK 19 guidelines.The assessment assigned points for features indicating a high risk of bias, ranging from 0 (no risk of bias) to 8 (high risk of bias) (Supplemental Information 3 in Extended data 13 ).Conflicts relating to risk of bias were assessed by the same arbitrator that arbitrated conflicts relating to data extraction (JL, FS).The risk of bias of studies included in one or more meta-analysis were additionally assessed using the QUIPS tool 20

Meta-analysis
We undertook a meta-analysis of associations between a biomarker and outcome if (i) the same biomarker was assessed in ≥2 studies, (ii) means/medians (and errors) of biomarkers were reported and (iii) the same outcome measure was reported between studies at the same time point after ICH onset.We used the metafor (RRID:SCR_003450) (version 3.0-2) and tidyverse (RRID:SCR_019186) packages in R Project for Statistical Computing (RRID:SCR_001905) (version 3.6.3) 21.Median and range were first converted to mean and standard deviation based on previously published methods 22 .We then calculated individual standardized mean differences (SMDs) using Hedges g, before estimating inter-study variability (τ 2 ) using a restricted maximum likelihood randomeffects model, generating a summary standardized mean difference.We used SMD since this was the most commonly reported outcome measure.However we also used the stata command metan to perform meta-analysis of odds ratios where they were provided.We assessed heterogeneity using the Higgins' I 2 statistic and Q statistics.Where more than 10 studies were included in meta-analysis, we evaluated outliers both graphically 23 and via influence diagnostics 24 and if we identified an outlier, sensitivity was assessed by leave-one-out analysis.Publication bias was evaluated graphically with funnel plots and Egger's regression test 25 where more than 10 studies were included in a meta-analysis, we used meta-regression to evaluate the influence of the pre-defined moderators ICH volume and age on the observed model variances.If more than 10 studies were included in meta-analysis, studies were stratified into high-quality (0-1 risk of bias score) and low quality (≥2 risk of bias score) according to the eight item composite scale and sub-group analysis performed to determine the impact of study quality on summary estimates.We have also stratified studies into low, moderate or high risk of bias according to the QUIPS rating tool and once again performed subgroup analyses to determine the impact of study quality on summary estimates.I 2 values of 0-39% were considered small, 40-69% moderate and 70-100% high.The code to reproduce the meta-analysis can be found at Extended data 26 .

Results
Our search yielded 8,794 articles, of which 8,615 were unique.In total, 98 studies of 50 unique inflammatory biomarkers in 18,000 participants were included in our narrative synthesis, of which 93 (n=17,702 participants) examined the association between at least one biomarker and clinical outcome.Overall, 11 of these also examined PHO (n=1,001 participants).Five studies (n=298) examined an association between biomarkers and PHO alone (Figure 1).From all 98 included studies, 85 (87%) studies reported biomarker levels on admission or within 24 hours from ICH onset, of which 13 (15%) also reported biomarkers at later time points.A total of 12 (12%) studies only reported biomarker levels at later time points ranging from <48 hours to 30 days after ICH onset, and one study did not report biomarker time point.Overall, 37 (38%) studies reported on a single biomarker, and the remainder reported on two biomarkers (n=21, 21%), three biomarkers (n=14, 14%) or four or more biomarkers (n=26, 27%).

Narrative synthesis of the associations between inflammatory biomarkers and perihaematomal oedema
In total, 16 studies 27-42 of 1,299 participants assessed the relationship between 21 inflammatory biomarkers and PHO (14 by CT, two by MRI) (Table 1; participant characteristics listed in Supplemental Table 3 13 ), of which we excluded one study 31 which assessed PHO using midline shift only.The median risk of bias was 2 [0-3].Due to broad heterogeneity in the method and timing of PHO measurements between studies coupled with variation in the ways in which studies assessed the relationship between a given biomarker and PHO, we could not perform a meta-analysis of the association between inflammatory mediators and PHO.Therefore, what follows is a narrative synthesis of all published articles that analysed the association between one or more inflammatory biomarkers and PHO (Table 1).
The most frequently measured biomarker was MMP-9, which was assessed in five studies (CT n=4 34-37 , MRI n=1 28 ; n=225), three [35][36][37] of which found that MMP-9 levels were positively associated with PHO volume in univariate analyses only.MMP-2 37 and MMP-8 34 were positively associated with absolute PHO volume.Comparisons of MMP-3 and PHO were inconsistent with one study finding no association 34 and one study showing a positive association 28 .TIMP-1 was not associated with PHO in two studies 35,37 , but had an inverse association with PHO volume in one 36 .
Higher levels of ferritin were associated with PHO growth 32 , relative 40 , or absolute PHO volume 29 .White blood cell (WBC) count had no association with relative PHO volume 39,40,42 , whereas neutrophil count and neutrophil-to-lymphocyte ratio (NLR) had a positive correlation with midline shift after adjusting for ICH volume and infection 31 .NLR was positively correlated with PHO in three studies 30,38,39 .IL-4 was positively associated with absolute PHO in one of two studies 41 .IL-6 was associated with absolute PHO volume in two studies 33,41 and TNFα was associated with absolute volume in univariate but not multivariate analyses 33 .
Adhesion molecules ICAM1 and VCAM1 were not associated with absolute PHO volume in multivariate analyses 33 .

Meta-analysis of the association between inflammatory biomarkers and clinical outcome
From 93 studies examining the association between inflammatory biomarkers and clinical outcome after ICH, the median risk of bias was two [1-3].The most commonly reported outcome measure was the modified Rankin scale (mRS) (n=59, 63%) and the most frequent time point for outcome ascertainment was 90 days (n=41, 44%) (Supplemental Figure 1A and 1B 13 ).Therefore, studies reporting mRS at 90 days were selected for meta-analysis with poor outcome defined as death or dependency (mRS 3-6).Overall, 20 studies of nine biomarkers were included in meta-analysis (Table 2).Of these 20 studies, only one study reported steroid use at the time of ICH 43 and no studies reported on the use of osmotic agents, NSAIDs, or other immunomodulatory drugs at the time of ICH.No studies reported the number of febrile patients on admission, but 13 reported the presence of intercurrent infection with 10 using it as an exclusion criterion.Seven studies reported the use of antiplatelet agents and nine studies reported anticoagulant medication at the time of ICH (Supplemental Table 4, available as Extended data 13 ).All 93 studies examining the association between inflammatory biomarkers and clinical outcome have been included in a narrative synthesis (Supplemental Table 2, available as Extended data 13 ).
From three studies reporting on ferritin and clinical outcome 29,32,105 , two studies of 155 participants measuring ferritin upon admission were included in a meta-analysis 29,32 .Neither of these studies included a multivariable analysis.There was no association between ferritin and death or dependency at 90 days (pooled SMD 1.59; 95% CI [-0.94, 4.13]; p=0.22) (Supplemental Figure 7 13 ).There was a high level of statistical heterogeneity (I 2 =98%) between studies.
Overall, 11 other cytokines were measured in 10 studies 33,34,72,101,102,104,[108][109][110][111] where high levels of seven cytokines were associated with worse clinical outcome, and this association was sustained in multi-variate analysis in four studies.
Two studies 64,93 of 300 participants reporting on HMGB1 upon admission and outcome were meta-analysed, neither of which included a multivariable analysis.Higher HMGB1 levels were associated with death or dependency at 90 days (pooled SMD 1.67; 95% CI [0.05, 3.30]; p=0.04) (Figure 5).There was a high level of heterogeneity (I 2 =95%) between studies.

Tissue remodelling factors.
Nine studies 28,34-37,72,103,118,119 reported on MMP-9 and clinical outcome, six of which found no association with outcome 34,36,37,72,104,119 .An additional six MMPs and two TIMPs were also reported by five of these studies 28,34,36,37,119 and no consistent association with clinical outcome was found.Meta-analysis was not performed because only single studies reported both biomarker concentrations and mRS at 90-days for each analyte.
Adhesion molecules.Eight adhesion molecules were reported alongside clinical outcome in seven studies 28,33,65,66,103,120,121 .However, as no single molecule was reported with 90-day mRS more than once, meta-analysis was not performed.Higher levels of four adhesion molecules; sCD40L, ICAM1, Selectin-E, Selectin-P, were associated with worse outcome in three studies 33,65,103 .Three molecules; VCAM1, VEGF, Tim-3, showed no association in four studies 28,33,66,121 .Lower amounts of CD163 were associated with poor outcome in one study 120 .

Narrative synthesis of the association between inflammatory biomarkers and both PHO and outcome
Overall, 11 studies of 1,001 participants measured both PHO and clinical outcome.We did not find any biomarker that was consistently associated with both PHO and clinical outcome in more than one study and thus no meta-analysis could be performed.In one study of 116 participants, higher levels of IL-6, TNFα and ICAM1 were associated with larger PHO volumes and worse clinical outcomes in univariable analyses 33 .In individual studies, S100B 27 and ferritin 29 were positively associated with both PHO and outcome in univariable analyses only.In one study (n=57) higher levels of MMP-9 were associated with larger PHO volumes  in adults with supratentorial deep ICH and worse clinical outcomes at days 3-6 after ICH onset in univariable analyses 35 .Studies of lymphocyte counts and NLR and their associations with PHO and outcome reported inconsistent findings 30,31 .

Discussion
This systematic review and meta-analysis is the first to assess the relationship between inflammatory biomarkers and both PHO and outcome following ICH.From 93 studies examining the relationship between 50 biomarkers and functional outcome after ICH, we found that higher circulating levels of two biomarkers; fibrinogen and HMGB1, were associated with death or dependency at 90 days after ICH.A higher level of CRP seemed to be associated with death or dependency after ICH but when odds ratios were metaanalysed, this association did not remain.A higher level of WBC also seemed to be associated with death or dependency but this association did not remain in a pooled subgroup analysis of studies with a low risk of bias.We did not find an association between neutrophil count, ferritin, IL-6, TNFα or S100B and clinical outcome.We did not find any biomarker that was associated with both PHO and outcome in more than one study.The methodological heterogeneity between studies of inflammatory biomarkers and PHO precluded meta-analysis.However, we have narratively synthesised the findings from all published studies investigating the relationship between circulating inflammatory mediators and PHO MMP-9 was the most frequently assessed biomarker in studies of PHO.It may be associated with PHO after ICH, but studies have been small with inconsistent findings and rarely controlled for important covariates such as ICH volume, which may act as confounders.There was heterogeneity in the timepoint of biomarker measurement which may also contribute to inconsistent findings since the level of a serum biomarker will vary in relation to the time of measurement after ICH onset.As MMPs have a role in the breakdown of the blood brain barrier, and contribute to vasogenic and cytotoxic PHO 122,123 , we believe this merits further investigation.Currently, these results are hypothesis generating and require replication in larger studies using a prespecified scanning protocol for the measurement of PHO.
A key finding from this study is the association between higher circulating HMGB1 levels and poorer outcome after ICH.HMGB1 is released from necrotic and inflammatory cells and stimulates DAMP receptors to amplify release of pro-inflammatory cytokines and recruit peripheral immune cells 124,125 .Its' activity is increased after ICH in humans and is associated with the development of PHO 126 .Since inhibiting HMGB1 improves outcome in animal models of ICH 127,128 it may be a potential therapeutic target that merits further investigation.
of ischemic stroke that have found similar associations 129 .CRP may be associated with poor outcome after ICH 130 , but we did not find a consistent association.
Although there seemed to be an association between WBC and outcome, the association no longer remained significant in a subgroup analysis of three studies (total sample size =304) with a low risk of bias.This may because we lacked power.This requires further exploration in larger prospective ICH cohorts.
These findings support the role of inflammation in ICH but do not provide information on the exact inflammatory pathways engaged or lead to potential therapeutic targets.Unlike HMGB1, which has been shown in preclinical studies to be specifically neutralised by anti-HMGB1 monoclonal antibodies 128 , CRP is a non-specific marker of inflammation.For example, CRP is produced in the acute phase of most forms of inflammation and is thus non-specific to the inflammatory response to ICH.We believe that the analysis of more specific inflammatory mediators is required in order to gain a deeper understanding of the immune response to ICH and identify potential therapeutic targets.Of note however, the association of CRP with poor outcome is stronger as the median age of the cohort decreases.Whilst this could be explained by CRP biology as background serum levels of CRP increase during ageing 131 , it could also be linked to confounders such as patient selection, since studies with lower median cohort ages had higher risk of bias scores.
Our meta-analysis did not find an association between circulating levels of either IL-6 or TNFα with death or dependency at 90 days.These cytokines stimulate the expression of acute phase proteins, such as CRP, and are thought to play detrimental roles in the pathogenesis of ICH 8,132 .Several factors may explain why we did not find an association: only four studies were included in these meta-analyses, the time point of biomarker measurement in relation to ICH onset varied, and peripheral circulating levels of these cytokines may not be comparable to concentrations in the brain.Moreover, the majority of studies only reported biomarker measurements upon admission or within 24hr.It is thus difficult to establish if these biomarker levels are reflective of the baseline physiology of individuals at risk of developing worse outcome or are directly caused by the onset of ICH itself.Longitudinal, serial blood sampling would assist in measuring the effect of ICH on systemic inflammatory levels independently of baseline levels.
We did not find any biomarker that was associated with both PHO and outcome in more than one study.Future studies should aim to identify if any biomarker is associated with both, since such a biomarker could either be a prognostic marker after ICH or act as an inflammatory mediator which would be more likely to be a potential therapeutic target.Since both animal studies and human studies (of brain tissue and serum) suggest the involvement of inflammatory pathways such as IL-1, HMGB-1, TNF in the perihaematomal region 133 and the development of PHO 126 , and PHO itself is associated with outcome after ICH 134,135 , it is plausible that one or more such biomarkers could be involved in the inflammatory response in the brain following ICH.' Modulating the neuroinflammatory response after ICH, for example by using an IL-1 receptor antagonist (NCT04834388) may therefore hold promise.
Our review is strengthened by a comprehensive search strategy of historic and contemporary literature without language or publication date restrictions.We did not limit our search strategy based on a priori knowledge of pre-specified biomarkers, and we identified many studies that were not considered in a recent review of the association between biomarkers and prognosis after ICH 136 .We believe this is the first systematic review to assess the relationship between inflammatory biomarkers and both PHO and functional outcome after ICH.Critical appraisal of studies was determined by two independent reviewers and most meta-analyses were not affected by publication bias.
This study has some limitations.Our risk of bias assessment used a summed score, which may not fully reflect the degree of bias in certain studies.We encountered a high level of heterogeneity between studies in relation to biomarker measurement and PHO assessment, which precluded meta-analyses of some biomarkers.Given the heterogeneity and small numbers of studies of some biomarkers, we were unable to use meta-regression to determine the influence of variables such as ICH volume 137 on the association of all biomarkers with clinical outcome.None of the studies included in meta-analysis reported whether the biomarker helped to improve prediction of prognosis after ICH when compared or added to existing prognostic scores for ICH and this should be examined in future studies.We were unable to explore how variation in PHO measurement between studies affects any association between PHO and prognosis after ICH but this has been discussed elsewhere 134,135 .

Conclusions
Higher levels of CRP, fibrinogen and HMGB1 were associated with worse outcome after ICH.Future prospective studies should prioritise the investigation of specific inflammatory mediators (such as HMGB1, cytokines, DAMPs and adhesion molecules) and adjust for key covariates such as ICH severity to better understand the pathophysiology of PHO and inflammation after ICH.This may reveal novel biomarkers, identify potential therapeutic targets and give better insights into certain immune-related post-ICH sequelae, such as infection.

Ethics approval
An ethics statement is not applicable as this study is exclusively based on published literature.

Extended data
Figshare: Supplementary Material to Kirby et al., 2023_Asso The only minor edit I would suggest is that in the Narrative synthesis of the association between inflammatory biomarkers and perihematomal edema section, the authors discuss a few studies that investigated the association between certain markers and midline shift, but this is the first time MLS is mentioned.In the Introduction and Methods sections they only discuss PHO as an imaging marker of inflammation.I think it could be reasonable to use midline shift as an imaging marker of inflammation, especially if they believe/state that the midline shift is due to PHO, but this is not discussed.If these studies investigated MLS as an imaging marker of neuroinflammation since it was due to PHO, then it would make sense to include these studies but the authors should make this clear.

Are sufficient details of the methods and analysis provided to allow replication by others? Yes
Is the statistical analysis and its interpretation appropriate?I cannot comment.A qualified statistician is required.Thank you for giving me the opportunity to comment on the manuscript "Association between circulating inflammatory biomarkers and functional outcome or perihaematomal oedema after ICH: a systematic review & meta-analysis".The authors hypothesize that inflammatory biomarkers in peripheral blood may be associated with clinical outcome following ICH, as well as perihaematomal oedema (PHO).The latter association was assessed in 16 studies while the former relation between biomarkers and clinical outcome was examined in 93 studies.The authors came to the conclusion that higher levels of WBC, CRP, fibrinogen and HMGB1 were associated with worse outcome after ICH.

II Strengths
Overall, this is a very comprehensive and excellently conducted meta-analysis which will hopefully act as an inspiration/stepping stone for further research into this important topic.
The study follows the PRISMA guidelines.The introduction clearly states what the aims of the study are and why this project was undertaken, followed by appropriate methods to achieve that aim.The results are well presented with forest plots, tables and additional material in the supplementary data.
As a reviewer, I have a few suggestions for consideration to the authors.I divided them into the section of the manuscript rather than major/minor points.

Methods
How was the de-duplication done?This would be a small but helpful information for readers and especially researchers who would like to embark on conducting a meta-analysis themselves.
○ I am missing a paragraph describing the outcome measures (and potentially timepoints) used in the studies.Since this study looks at associations between biomarkers and outcome, this information is crucial.Please add this.

○
Using standardized mean differences (SMD) in this analysis seems very appropriate to me.But it would be important to give the reader a brief explanation as to why you picked this measure.Also, mention which SMD you used (Cohen's d, Hedges, Glass' delta etc., see Andrade, J Clin Psychiatry 2020)

○
Please provide a sentence or two as to why you chose the restricted maximum likelihood random effects model over other very commonly used models such as the DerSimonian-Laird model.

○
Risk of bias assessment: The authors developed a new scale to assess the quality of the included studies.Considering the large number of papers included in this study it is understandable that the authors would aim for a simple and short form of assessment.That being said, it is important to assess all aspects of a study and it seems like there are no questions pertaining to statistical analyses and/or confounding in the self-developed questionnaire.Please provide a reasoning for the decision to use a self-developed scale (instead of one of the existing ones such as STROBE-ME or QUIPS in their entirety) and how you picked the final items for your scale.Also, add the initials of the authors who performed the risk of bias and explain how you resolved potential disagreements.

Results:
On page 9 it would be helpful for the reader to mention the term meta-regression in the text ○ You report after each meta-analysis that the heterogeneity between studies was high.It would be helpful to have the I 2 in a bracket provided to the reader (the actual number).

○
You mention the sensitivity analysis in the methods section, but I don't see any results to this effect for any of the meta-analyses or any mention of it except of for the fibrinogen group.Was this done for all meta-analyses which had more than 3 studies included?

○
In the paragraph about Cytokines (page 11), you provide the pooled estimates for the analysis and for the analysis sans outlier and mention they were similar.It seems that they were not exactly similar, but they both stayed non-significant.

Are sufficient details of the methods and analysis provided to allow replication by others? Yes
Is the statistical analysis and its interpretation appropriate?Yes

Are the conclusions drawn adequately supported by the results presented in the review? Yes
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Intracerebral hemorrhage, stroke, I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
suggestions for consideration to the authors.I divided them into the section of the manuscript rather than major/minor points.III Suggestions to improve the manuscript Methods How was the de-duplication done?This would be a small but helpful information for readers and especially researchers who would like to embark on conducting a meta-analysis themselves.
Response: Thank you for requesting this clarification.We have amended the methods: 'After automated de-duplication in Covidence (covidence.org),three authors…'

I am missing a paragraph describing the outcome measures (and potentially timepoints) used in the studies. Since this study looks at associations between biomarkers and outcome, this information is crucial. Please add this.
Response: Thank you for requesting this.All measures of outcome and time points are given in Supplemental Table 1.
Using standardized mean differences (SMD) in this analysis seems very appropriate to me.But it would be important to give the reader a brief explanation as to why you picked this measure.Also, mention which SMD you used (Cohen's d, Hedges, Glass' delta etc., see Andrade, J Clin Psychiatry 2020) Response: Thank you for requesting this clarification.We used the Hedges g for estimation of SMD because it considers sample size when calculating overall effect size and have now included this in the manuscript in the Methods section (meta-analysis).We used SMD because it was the most commonly used outcome measure.

Please provide a sentence or two as to why you chose the restricted maximum likelihood random effects model over other very commonly used models such as the DerSimonian-Laird model.
Response: Thank you for requesting this clarification.We chose the restricted maximum likelihood random effects model because it shows slightly less bias in simulations ( https://onlinelibrary.wiley.com/doi/10.1002/jrsm.1316[onlinelibrary.wiley.com]), is easy to access through openly available R packages such as metafor, and computational burden was not a factor.Response: Thank you for requesting this clarification.As per our response to a comment raised by reviewer 1, we used a self-developed scale as although the STROBE-ME guidelines, Newcastle Ottawa scale and REMARK guidelines cover aspects of assessment of risk of bias we did not think that any of these three guidelines covered all the aspects that should be included (for example -the STROBE-ME and Newcastle Ottawa scale checklists do not include any aspects specific to biomarker measurement, while the REMARK checklist seems to place less emphasis on participant selection in comparison to the others but includes specific information relevant to serum biomarker studies.)We have now also rated each of the studies included in a meta-analysis using the QUIPS tool and included this information in the supplement.We have also made it clearer in the methods (data extraction section) that risk of bias was assessed by the same authors who extracted data from the studies (CK, JB or NS) and conflicts relating to RoB were assessed by the same arbitrator that arbitrated conflicts relating to data extraction (JL, FS).We have amended the abstract to indicate that we assessed risk of bias using both a composite scale and the QUIPS tool (for those studies being meta-analysed).

Results: On page 9 it would be helpful for the reader to mention the term metaregression in the text
Response: Thank you.We have added the term 'meta-regression' into the text.
You report after each meta-analysis that the heterogeneity between studies was high.It would be helpful to have the I2 in a bracket provided to the reader (the actual number).
Response: Thank you for requesting this clarification.We have added the I 2 statistic into each sentence in the results section which describes heterogeneity.
You mention the sensitivity analysis in the methods section, but I don't see any results to this effect for any of the meta-analyses or any mention of it except of for the fibrinogen group.Was this done for all meta-analyses which had more than 3 studies included?
Response: Thank you for this important comment.We only used a sensitivity analysis if we identified an outlier.We have made this clearer in the methods (meta-analysis section) 'Where more than three studies were included in meta-analysis, we evaluated outliers both graphically and via influence diagnostics and if we identified an outlier, sensitivity was assessed by leave-one-out analysis.
In the paragraph about Cytokines (page 11), you provide the pooled estimates for the analysis and for the analysis sans outlier and mention they were similar.It seems that they were not exactly similar, but they both stayed non-significant.
Response: Thank you.We have corrected this in the results section (cytokines).'However, pooled estimates remained non-significant when this study was removed…' In the same chapter as above on page 11 you mention 11 other Cytokines.It was unclear to me as to why there wasn't any meta-analysis on this -especially since you mention a significant association with outcome.Could you provide some clarification to this paragraph?
Response: Thank you for requesting this clarification.Apart from IL-6 and TNFa, no other cytokines were analysed with the same outcome measure in more than one study so no meta-analysis could be performed (Supplemental Table 1).Response: Thank you for this important point.Since funnel plots should only be used when there are at least 10 studies included in the meta-analysis (https://handbook-5-1.cochrane.org/chapter_10/10_4_3_1_recommendations_on_testing_for_funnel_plot_asymmetry.htm ) we have now removed the funnel plots for biomarkers with fewer than 10 studies (fibrinogen, IL-6, S100B).We have provided a new DOI for supplemental material.Discussion: In the first sentence of the discussion you talk about "all" biomarkers.Clarify whether you mean all existing ones to date (in all literature) or all you found with regards to the question you asked.

Funnel plots in
Response: Thank you for requesting this clarification.We meant all existing ones to date identified in the literature by our search strategy, but to simplify the wording we have removed the word 'all'.
It seems that PHO and biomarkers are somewhat treated as separate entities in this manuscript, however, both are interrelated and part of the secondary inflammatory response to the ICH.It would be helpful to work out/clarify this relationship a bit more in the discussion and/or conclusion and try to elaborate how they both relate to outcome.The conclusion is very general.As reader I would be curious to know -after this extensive presentation and summary of literature -how this all translates into practice.What would we expect if there was a clear relationship between biomarker and outcome or a clear relationship between PHO and biomarkers/outcome.What could potential therapeutic targets be or how could a potential therapy look like?How would this change clinical practice, if at all?
Response: Thank you for both these comments.We have added the following paragraph into the discussion: 'Future studies should aim to identify if any biomarker is associated with both, since such a biomarker could either be a prognostic marker after ICH or act as an inflammatory mediator which would be more likely to be a potential therapeutic target.Since both animal studies and human studies (of brain tissue and serum) suggest the involvement of inflammatory pathways such as IL-1, HMGB-1 and TNF in the perihaematomal region 130 and the development of PHO, 123 and PHO itself is associated with outcome after ICH, 131,132 it is plausible that one or more such biomarkers could be involved in the inflammatory response in the brain following ICH'.Modulating the neuroinflammatory response after ICH, for example by using an IL-1 receptor antagonist (NCT04834388) may therefore hold promise'.
Competing Interests: No competing interests were disclosed.
The authors categorized inflammatory markers into six groups based on their biological properties: immune cells, cytokines/chemokines, tissue remodeling factors, damageassociated molecular patterns (DAMP), acute phase reactants, and adhesion molecules.However, the classification criteria may need reconsidering because some categories, such as acute phase reactants, may overlap.The authors should justify their classification system.

3.
The authors studied the connection between inflammatory markers and the prognosis of ICH, but they did not examine the relationship between PHO and ICH prognosis.The issue may be that various studies used different methods to measure PHO, so the differences in research approaches require clarification and summarization.The authors should investigate whether variations in research methods for measuring PHO could impact the prognosis of ICH.

4.
In their study, the authors found that MMP-9 showed a positive correlation with the volume of PHO and that MMP-2 and MMP-8 were positively correlated with the absolute volume of PHO.However, another study revealed that on the 7th day after ICH, MMP-2 had a negative correlation with PHO volume [3] .It is worth noting that the authors did not consider changes in the inflammatory response at different time points, which could affect the interpretation of the results.It is expected that there would be differences in each inflammatory marker at various stages of ICH.

5.
According to the study, there is a connection between the volume of PHO and IL-10.However, previous research has stated that there is a negative relationship between PHO volume and IL-10 [4] .Another study found that higher levels of IL-10 in the blood are linked to a favorable prognosis 90 days after ICH.Additionally, administering IL-10 through the nose has decreased PHO and enhanced functional outcomes of mice with ICH [5] .As a result, the authors need to approach their findings with caution.

Are the conclusions drawn adequately supported by the results presented in the review? Partly
Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Intracerebral hemorrhage, Cerebrovascular disease
We confirm that we have read this submission and believe that we have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however we have significant reservations, as outlined above.
Thank you for requesting this clarification.We list the criteria for determining which studies were deemed ineligible in Figure 1, with the most frequent five reasons that studies were deemed ineligible being: Studies were published as conference abstracts only 1.
Studies were narrative or systematic reviews 2.
No relationship between an inflammatory biomarker and either PHO or clinical outcome was measured.

3.
Patients treated with a surgical intervention included in the analysis 4.
Mixed causes of ICH or stroke included in the analysis.5.
We excluded studies that included patients treated with a surgical intervention (when we were unable to separate in the analysis those treated with surgery from those not treated with surgery) since surgery may affect both the biomarker (for example, CRP may rise with surgery) and PHO or outcome, thereby acting as a confounder.We excluded studies which had a mixed cohort (for example including patients with ischaemic stroke or patients who had ICH due to a structural cause such as an aneurysm or a tumour) when we were unable to separate those with spontaneous ICH from the rest in the analysis.The inflammatory response after spontaneous ICH may differ from those with ICH or stroke due to another cause.We agree that it is important to ensure that the quality of the selected studies meets rigorous research standards and have therefore re-rated all studies included in metaanalyses using the QUIPS tool (see response to reviewer 1) whilst also retaining our risk of bias assessment using the composite scale.
In the Introduction, the authors point out that PHO is an imaging biomarker of the neuroinflammatory response.However, it's important to note that PHO is related to neuroinflammation and other secondary brain injury after ICH, including blood-brain damage [1].Furthermore, PHO can be divided into different stages, and neuroinflammation may lead to PHO development rather than PHO solely representing neuroinflammation [2].
Response: Thank you for raising this point.We have added in text to reflect this in the introduction -Perihaematomal oedema (PHO) is thought to be an imaging biomarker of the neuroinflammatory response 5 and neuroinflammation may lead to PHO development.PHO is thought evolve in stages with initial cytotoxic oedema leading to vasogenic oedema.
The authors categorized inflammatory markers into six groups based on their biological properties: immune cells, cytokines/chemokines, tissue remodeling factors, damage-associated molecular patterns (DAMP), acute phase reactants, and adhesion molecules.However, the classification criteria may need reconsidering because some categories, such as acute phase reactants, may overlap.The authors should justify their classification system.
Response: Thank you for requesting this clarification.We have modified the text (Methods, eligibility criteria section): 'We selected biomarkers if they were a marker (e.g., C-reactive protein) or a mediator (e.g., cytokines) of the inflammatory response and grouped the inflammatory biomarkers into six categories based on broad biological activity.Categories were defined based on discussions between domain experts (JB, BM).…immune cells, acute phase reactants (defined here as liver-produced plasma proteins (PMID: 21430962) cytokines/chemokines, damage-associated molecular patterns (DAMPs) (defined here as molecules released from intracellular compartments during inflammation (PMID: 25391648), tissue remodelling factors and adhesion molecules.
The authors studied the connection between inflammatory markers and the prognosis of ICH, but they did not examine the relationship between PHO and ICH prognosis.The issue may be that various studies used different methods to measure PHO, so the differences in research approaches require clarification and summarization.The authors should investigate whether variations in research methods for measuring PHO could impact the prognosis of ICH.
Response: Thank you for this important comment.We did not examine how differences in the measurement of PHO between studies affected prognosis after ICH since this systematic review focused upon the relationship firstly, between serum biomarkers and PHO and secondly, between serum biomarkers and outcome rather than the relationship between PHO and outcome.However this has been discussed in other systematic reviews and meta analyses (Marchina et al. 2023: 10.1016/j.jstrokecerebrovasdis.2023.107204;Cliteur et al. 2023: 10.1177/23969873231157884) and we recognise this as a limitation: 'We were unable to explore how variation in PHO measurement between studies affects any association between PHO and prognosis after ICH but this has been discussed elsewhere' (Discussion) In their study, the authors found that MMP-9 showed a positive correlation with the volume of PHO and that MMP-2 and MMP-8 were positively correlated with the absolute volume of PHO.However, another study revealed that on the 7th day after ICH, MMP-2 had a negative correlation with PHO volume [3].It is worth noting that the authors did not consider changes in the inflammatory response at different time points, which could affect the interpretation of the results.It is expected that there would be differences in each inflammatory marker at various stages of ICH.
Response: Thank you for this comment.We now recognise this in the discussion: 'There was heterogeneity in the timepoint of biomarker measurement which may also contribute to inconsistent findings since the level of a serum biomarker will vary in relation to the time of measurement after ICH onset.' According to the study, there is a connection between the volume of PHO and IL-10.However, previous research has stated that there is a negative relationship between PHO volume and IL-10 [4].Another study found that higher levels of IL-10 in the blood are linked to a favorable prognosis 90 days after ICH.Additionally, administering IL-10 through the nose has decreased PHO and enhanced functional outcomes of mice with ICH [5].As a result, the authors need to approach their findings with caution.
Response: Thank you for this important comment.We included reference 4, which the reviewer mentions above, in table 1, 'Associations between circulating inflammatory biomarkers and perihaematomal oedema' but incorrectly listed that it showed a positive correlation between IL-10 and PHO volume.This error may in part be due to the original paper incorrectly stating in the conclusion of the abstract that 'serum levels of…..IL-10 are positively correlated.' Thank you for pointing out that it actually demonstrated a negative relationship and we have corrected this both in table 1 and in the results section (where we have removed the reference to IL-10 in the sentence below): 'IL-4 was positively associated with absolute PHO in one of two studies 38 .advisable to analyse hazard ratios, odds ratios, and risk ratios separately, as well as unadjusted and adjusted associations separately, instead of the reported standardised mean differences.

Minor points
Incremental  Response: Thank you for requesting this clarification.We used a self-developed scale as although the STROBE-ME guidelines, Newcastle Ottawa scale and REMARK guidelines cover aspects of assessment of risk of bias we did not think that any of these three guidelines covered all the aspects that should be included (for example -the STROBE-ME and Newcastle Ottawa scale checklists do not include any aspects specific to biomarker measurement, while the REMARK checklist seems to place less emphasis on participant selection in comparison to the others but includes specific information relevant to serum biomarker studies.)We have now also rated each of the studies included in a meta-analysis using the QUIPS tool and included this information in Supplemental Table 1 the extended data.In accordance with Riley et al. (doi: 10.1136/bmj.k4597.)We selected a priori which domains we considered most important (study attrition, study participation and study confounding) and assigned each study an overall risk of bias based chiefly upon these domains.12/20 (60%) had a moderate risk of bias, five studies (25%) had a low risk of bias and the remaining three (15%) had a high risk of bias.In the original submission if there were more than 10 studies included in meta-analysis, studies we stratified them into high-quality (0-1 risk of bias score) and low quality (≥2 risk of bias score) according to the eight item composite scale and performed sub-group analysis to determine the impact of study quality on summary estimates.We have kept these analyses in the submission but have now also re-run our analyses of (1) the association between WBC and outcome, and (2) the association between CRP and outcome having stratified the studies by their risk of bias measured by the QUIPS tool (low/moderate/high) to determine the impact of study quality on the association between the biomarker and outcome.We have mentioned this in the methods section, page 5.

Analysis of WBC and outcome:
We have inserted a figure 2B into the supplement and inserted the following text into the results section, 'immune cells': Study quality did not influence pooled associations when risk of bias was assessed by the composite scale (p=0.07)(Supplemental Figure 2A 13 ) but when assessed by the QUIPS tool the association between WBC and outcome was no longer statistically significant in studies with a low risk of bias (pooled SMD 0.22; 95% CI -0.04-0.48).

Analysis of CRP and outcome:
We have inserted a figure 5B into the supplement and inserted the following text into the results section, 'acute phase reactants': 'Study quality when assessed both by the composite scale and by the QUIPS tool did not influence pooled associations (Supplemental Figure 5A and Figure 5B).'We have also adjusted the abstract and the discussion accordingly to reflect the association between WBC and outcome no longer being significant in a meta-analysis of high quality studies.
Meta-analysis: Following the recommendations of Riley et al. (BMJ, 2019) 3, it would be advisable to analyse hazard ratios, odds ratios, and risk ratios separately, as well as unadjusted and adjusted associations separately, instead of the reported standardised mean differences.
Response: Thank you for this comment.We selected studies for meta-analysis based upon those studies which had assessed the same biomarker and measured outcome in the same way at the same time point (Methods, meta-analysis section, page 3).Of those studies which contributed to any of the meta-analyses, none provided a hazard ratio.Only single studies reported odds ratios of the association between S100b and outcome, HMGB1 and outcome and fibrinogen and outcome.No studies of ferritin or TNFa provided an odds ratio.We have now included a sentence in the methods section: 'We also used the stata command metan to perform meta-analysis of odds ratios where they were provided.'Only one study provided an adjusted odds ratio (OR) but five studies provided an unadjusted odds ratio of the association between WBC and outcome and we have now included a meta-analysis of these studies (Results, immune cells section).'The pooled OR of five studies 46,61 69,,81,86 which reported an unadjusted odds ratio was 1.06 (95% CI 1.02-1.10)'4 studies provided an adjusted odds ratio of the association between CRP and outcome.6 studies provided an unadjusted odds ratio of the association between CRP and outcome.We have added this information into the Results, acute phase reactants section.'The pooled adjusted OR of four studies, three of which had a low risk of bias according to QUIPS was 0.99 (95% CI 0.98-1.01).The pooled unadjusted OR of six studies, four of which had a low risk of bias according to QUIPS was 1.00 (95% CI 0.98-1.01).'We have amended the discussion to reflect the result of the pooled analysis of ORs for the association between CRP and outcome.'CRP may be associated with poor outcome after ICH, 127 but we did not find a consistent association.' 2 studies provided an adjusted OR for the association between neutrophils and outcome and this information has been added (Results, immune cells section).'A pooled OR of two studies that provided an adjusted OR showed no association between neutrophil count and outcome (OR 0.98, 95% CI 0.95-1.01).' 1 study [Mrackova 2020] reported an unadjusted OR for the association between IL-6 and outcome and one study provided an adjusted OR for the association between IL-6 and outcome.This information has been added to the text (Results, cytokines/chemokines section).Response: Thank you for this important point.We have added this into the Discussion, Limitations section.'None of the studies included in meta-analysis reported whether the biomarker helped to improve prediction of prognosis after ICH when compared or added to existing prognostic scores for ICH and this should be examined in future studies'.

Minor points
It would also be valuable if authors could report on the rate of studies adjusting for relevant confounding factors.At the very minimum, a valid association between a biomarker and poor outcome should be independent of age and clinical ICH severity.
Response: Thank you for this comment.We have included a sentence in each section commenting on this-WBCs: One of these studies (7.7%) included a multivariable analysis Neutrophil count: One study included a multivariable analysis CRP: Five of these studies (45%) included a multivariable analysis.Fibrinogen: One study included a multivariable analysis Ferritin: Neither study included a multivariable analysis IL-6: Two studies (50%) included a multivariable analysis TNFa: No studies included a multivariable analysis S100B: Two studies (50%) included a multivariable analysis HMGB1: Neither study included a multivariable analysis

Figure 1 .
Figure 1.Flow diagram of study selection for inclusion in systematic review and meta-analysis including the rationale for exclusion.PHO, perihaematomal oedema; ICH, intracerebral haemorrhage.
as mean (SD) or media [IQR] NR = Not Reported

Figure 2 .
Figure 2. Forest plot of pooled associations of WBC with death or dependency 90 days after ICH.Pooled association of circulating WBC with death or dependency at 90 days.Poor outcome defined as mRS 3-6.WBC, white blood cell; ICH, intracerebral haemorrhage; mRS, modified Rankin Scale; SMD, standardised mean difference.

Figure 3 .
Figure 3. Pooled associations of CRP with death or dependency at 90 days after ICH.A) Forest plot of pooled associations of circulating CRP with death or dependency at 90 days.Poor outcome defined as mRS 3-6.B) Influence of age on the association of CRP with death or dependency at 90 days after ICH.Study cohorts with lower median cohort age have larger standardized mean differences between CRP and death or dependency at 90 days after ICH.CRP, C-reactive protein; ICH, intracerebral haemorrhage; mRS, modified Rankin Scale; SMD, standardised mean difference.

Figure 4 .
Figure 4. Forest plot of pooled associations of IL-6 and TNFα with death or dependency 90 days after ICH.A) Pooled association of circulating IL-6 levels with death or dependency at 90 days.B) Pooled association of circulating TNFα levels with death or dependency at 90 days.Poor outcome defined as mRS 3-6.ICH, intracerebral haemorrhage; mRS, modified Rankin Scale; SMD, standardised mean difference.

Figure 5 .
Figure 5. Forest plot of pooled association of HMGB1 with death or dependency at 90 days after ICH.Pooled association of circulating HMGB1 with death or dependency at 90 days.Poor outcome defined as mRS 3-6.HMGB1, high mobility group box protein 1; ICH, intracerebral haemorrhage; mRS, modified Rankin Scale; SMD, standardised mean difference.
Sarah MarchinaDepartment of Neurology, Stroke Division, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts, USA I Summary

:
The authors developed a new scale to assess the quality of the included studies.Considering the large number of papers included in this study it is understandable that the authors would aim for a simple and short form of assessment.That being said, it is important to assess all aspects of a study and it seems like there are no questions pertaining to statistical analyses and/or confounding in the self-developed questionnaire.Please provide a reasoning for the decision to use a self-developed scale (instead of one of the existing ones such as STROBE-ME or QUIPS in their entirety) and how you picked the final items for your scale.Also, add the initials of the authors who performed the risk of bias and explain how you resolved potential disagreements.
the supplementary material: Every single funnel plot figure caption says: "Symmetrical distribution indicates a low probability of publication bias", however, Figure 4, 7 and 8 don't look very symmetrical which is probably why you reanalyzed without the outlier.I suggest to rephrase the figure caption for those figures.
Interests: No competing interests were disclosed.Reviewer Report 09 August 2023 https://doi.org/10.21956/wellcomeopenres.21268.r62058© 2023 Montellano F. This is an open access peer review report distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Felipe A. Montellano 1
Institute of Clinical Epidemiology and Biometry, University of Würzburg, Würzburg, Germany 2 Department of Neurology, University Hospital Würzburg, Würzburg, Germany Thank you for the opportunity to comment on the manuscript by Kirby et al. on the association between blood-based inflammatory biomarkers and functional outcome and/or perihaematomal oedema after intracerebral haemorrhage.In their systematic review and meta-analysis, they summarize the results of 16 publications examining the association between 21 inflammatory biomarkers and perihaematomal oedema, and 93 studies examining the association between inflammatory biomarkers and clinical outcome.They conclude that higher circulating levels of WBC, CRP, fibrinogen and HMGB1 are associated with poorer outcomes after ICH.Authors are to be congratulated for the thoroughness of their work and for embarking on the always difficult task of summarizing the very heterogeneous evidence that abounds in the field of biomarkers and prognosis research.In my opinion, some points merit further discussion:Risk of bias: Authors assessed risk of bias using a self-developed scale with elements from the Newcastle-Ottawa Assessment Scale, STROBE-ME and REMARK guidelines.While some relevant sources of bias are addressed by the proposed scale, others are not (particularly confounding, attrition, and statistical analysis).Further, as the authors rightly acknowledge in their limitations, a summed score probably does not fully reflect the risk of bias of certain studies.The QUIPS tool (Hayden JA et al.Ann Intern Med, 2013 1 ) was specifically designed to assess the risk of bias in prognostic factor research.The authors may want to use this tool in their systematic review, at the very least for those articles undergoing meta-analysis.This might change the interpretation of some of the findings of the manuscript.As an example, most of studies reporting on the association between white blood cell count and functional outcome after ischaemic stroke (Montellano FA et al.Stroke, 2021 2 ) arise from retrospective, convenience samples with suboptimal statistical modelling, lack of adjustment for relevant confounders and poor rates of follow up.1.Meta-analysis: Following the recommendations of Riley et al. (BMJ, 2019) 3 , it would be 2.
the opportunity to comment on the manuscript by Kirby et al. on the association between blood-based inflammatory biomarkers and functional outcome and/or perihaematomal oedema after intracerebral haemorrhage.In their systematic review and meta-analysis, they summarize the results of 16 publications examining the association between 21 inflammatory biomarkers and perihaematomal oedema, and 93 studies examining the association between inflammatory biomarkers and clinical outcome.They conclude that higher circulating levels of WBC, CRP, fibrinogen and HMGB1 are associated with poorer outcomes after ICH.Authors are to be congratulated for the thoroughness of their work and for embarking on the always difficult task of summarizing the very heterogeneous evidence that abounds in the field of biomarkers and prognosis research.Thank you In my opinion, some points merit further discussion: Risk of bias: Authors assessed risk of bias using a self-developed scale with elements from the Newcastle-Ottawa Assessment Scale, STROBE-ME and REMARK guidelines.While some relevant sources of bias are addressed by the proposed scale, others are not (particularly confounding, attrition, and statistical analysis).Further, as the authors rightly acknowledge in their limitations, a summed score probably does not fully reflect the risk of bias of certain studies.The QUIPS tool (Hayden JA et al.Ann Intern Med, 2013 1) was specifically designed to assess the risk of bias in prognostic factor research.The authors may want to use this tool in their systematic review, at the very least for those articles undergoing meta-analysis.This might change the interpretation of some of the findings of the manuscript.As an example, most of studies reporting on the association between white blood cell count and functional outcome after ischaemic stroke (Montellano FA et al.Stroke, 2021 2) arise from retrospective, convenience samples with suboptimal statistical modelling, lack of adjustment for relevant confounders and poor rates of follow up.

White blood cell count, Lymphocyte count, Neutrophil count, NLR
*Variables included in multivariate analysis:

Table 2 . Information on the 20 studies that were included in a meta-analysis of inflammatory biomarkers with 90-day mRS
. mRS, modified Rankin Scale; ICH, intracerebral haemorrhage; CRP, C-reactive protein; S100B, S100 Calcium Binding Protein B; WBC, white blood cell; HMGB1, high mobility group box protein 1; GCS, Glasgow Coma Scale.

have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard. Version 1
13ation between circulating inflammatory biomarkers and functional outcome or perihaematomal oedema after ICH.https://doi.org/10.6084/m9.figshare.2199567513.

Are the rationale for, and objectives of, the Systematic Review clearly stated? Yes
It seems that PHO and biomarkers are somewhat treated as separate entities in this manuscript, however, both are interrelated and part of the secondary inflammatory response to the ICH.It would be helpful to work out/clarify this relationship a bit more in the discussion and/or conclusion and try to elaborate how they both relate to outcome.What could potential therapeutic targets be or how could a potential therapy look like?How would this change clinical practice, if at all?
○In the same chapter as above on page 11 you mention 11 other Cytokines.It was unclear to me as to why there wasn't any meta-analysis on this -especially since you mention a significant association with outcome.Could you provide some clarification to this paragraph?○Funnelplots in the supplementary material:Every single funnel plot figure caption says: "Symmetrical distribution indicates a low probability of publication bias", however, Figure4, 7 and 8 don't look very symmetrical which is probably why you reanalyzed without the outlier.I suggest to rephrase the figure caption for those figures.○○Theconclusion is very general.As reader I would be curious to know -after this extensive presentation and summary of literature -how this all translates into practice.What would we expect if there was a clear relationship between biomarker and outcome or a clear relationship between PHO and biomarkers/outcome.

Are the rationale for, and objectives of, the Systematic Review clearly stated? Yes Are sufficient details of the methods and analysis provided to allow replication by others? Partly Is the statistical analysis and its interpretation appropriate
? I cannot comment.A qualified statistician is required.
value: Probably one of the most relevant questions in the field of prognosis research is whether novel prognostic factors, such as biomarkers, provide any incremental value on top of an established prognostic score , which often contains a set of readily available clinical variables (Steyerberg et al.PLoS Med.20134).It would be valuable if authors could report whether incremental value was investigated in the identified studies, either as improvement in discrimination, calibration, or overall model performance.Even if none of the identified studies addressed this issue, this would be a valuable insight.1.It would also be valuable if authors could report on the rate of studies adjusting for relevant confounding factors.At the very minimum, a valid association between a biomarker and poor outcome should be independent of age and clinical ICH severity.