Laboratory biomarkers of delayed cerebral ischemia following subarachnoid hemorrhage: A systematic review

Search strategy and selection criteria

PubMed and Embase were searched for relevant studies from January 1, 1985, to October 13, 2021. The search strategy can be found in the supplemental data (Supplementary material, page 2-3). Two authors (MAT and HL) independently selected eligible studies. Titles and abstracts were screened for potential eligibility when covering primary studies on biomarkers of DCI in SAH patients either written in Dutch or English. Of the selected studies, full texts were screened and studies were considered eligible if they met the following criteria: (1) patients with SAH of whom (2) at least 95% were diagnosed with a causative aneurysm, (3) measurement of at least one laboratory biomarker (4) during hospital admission, (5) using either the radiological or clinical definition of DCI as proposed in 2010 (criteria listed in Tab. I), and (6) the studies had to perform, or allow for, statistical analysis on the biomarker in question and DCI. Exclusion criteria were: (1) studies with mixed clinical and radiological DCI cohorts, in which patients with either clinical or radiological DCI could not be distinguished from the mixed cohort, (2) biomarker levels in patients with and without DCI were inconsistent (e.g., different numbers in tables/figures and text), (3) no level of statistical significance was given, nor could be calculated, (4) studies in children only, (5) the same patient population (similar years and hospital of patient inclusion) was used by more than one study and, within those studies, the same biomarker was measured and compared similarly between DCI and no DCI patients. In this case, the study with the smallest sample size, and if sample sizes were equal the most outdated study, was excluded. Lastly, we checked studies on nimodipine administration. In 1988 and 1989, two randomized controlled trials showed efficacy of nimodipine in preventing DCI, resulting in a shift in SAH management (5,6). Therefore, we excluded studies with SAH patients before 1985. SAH cohorts between 1985 and 1990 were only included if nimodipine administration was explicitly mentioned. All papers after 1990 were included, unless nimodipine was described explicitly as being not part of the treatment protocol.

Data extraction

The same authors (MAT and HL) independently extracted baseline SAH characteristics (age, gender, World Federation of Neurological Societies [WFNS] grade and Fisher grade), the study’s definition of DCI, the number of patients with and without DCI, and biomarker-related data.

If studies addressed both clinical and purely radiological DCI separately, biomarker-related data were extracted for both definitions. The biomarker-related data concerned laboratory biomarkers only (clinical and radiological biomarkers [heart rate, blood pressure, temperature, TCD velocities, computed tomography (CT) perfusion variables etc.] were left out of consideration). Biomarker data had to concern levels of individual biomarkers or ratios of two biomarkers, that is, combinations of three or more biomarkers, such as decision tree analysis or a self-made calculated score based on multiple biomarkers, were left out of consideration. Data included the biomarker medium (blood, cerebrospinal fluid [CSF], microdialysate [MD], urine, or other), timing of measurement(s), and all statistical analyses regarding the biomarker and the occurrence of DCI. If the biomarker medium was not mentioned, it was assumed to be blood. A biomarker measured in multiple media, for example, in blood and CSF, was considered as two separate biomarkers.

Nongenetic biomarkers were stratified into biomarker groups, which were based on the multifactorial pathophysiology of DCI as described by a previously published systematic review (7). Of biomarkers with multiple functions, we aimed to select the function which the authors of the study considered as DCI-related and subsequently categorized the biomarker into that specific biomarker group. Nongenetic biomarkers which could not be stratified to any biomarker group related to the pathophysiology of DCI were categorized as “other.” Within the “other” group, we aimed to cluster biomarkers with similar functions. Genetic biomarkers were categorized as a separate group. The extracted data were compared and discrepancies were discussed with a third author (DV) and solved.

Quality assessment

Two authors (MAT and DV) independently assessed the quality of included studies using the Newcastle-Ottawa Scale (NOS) for case control and cohort studies and the Cochrane Collaboration’s risk of bias assessment tool for randomized trials. Studies could obtain a maximum of eight or nine points, depending on the use of a nonexposed cohort. The Cochrane Collaboration’s risk of bias assessment tool comprises seven criteria, which can be scored as high, low, or unclear risk of bias. Discrepancies were discussed with a third opinion (WPV) and solved.

Data interpretation and analysis

Biomarkers of clinical DCI (primary outcome) were analyzed separately from biomarkers of radiological DCI (secondary outcome). We categorized biomarkers into significant and nonsignificant biomarkers. Biomarkers, which were studied by one study only, were considered significant if at least one measurement was significantly associated with DCI (p < 0.05) and nonsignificant if all measurements yielded nonsignificant results. If a biomarker was investigated by more than one study, the classification into significant and nonsignificant was either based on a biomarker-specific meta-analysis, if methodologically possible (similar type of analyses and timing of measurement), or by descriptive comparison. Meta-analyses were performed only if all relevant studies, which investigated the biomarker in question on a specific point in time after SAH, could be included, using Review Manager 5.4. For continuous data we used the inverse variance method with random effects model to calculate weighted mean differences with 95% confidence intervals. For dichotomous data we used the Mantel-Haenszel method with random effects model to calculate estimated odds ratio with 95% confidence intervals. A pooled effect size was estimated for each specific timing of biomarker measurement. Descriptive comparison of the studies was done by summarizing the results of all relevant studies in a biomarker-specific figure, which was then used to classify the biomarkers into significant, inconclusive, and nonsignificant. The biomarker-specific figures display the author, year, number of patients with and without DCI, the type of analyses, and the results divided into categories of sampling timing (day 0-3, 4-8, 9-13, and >14 after SAH). The method used for classification of biomarkers into significant, inconclusive, and nonsignificant based on the biomarker-specific figures is explained in the supplemental data (Supplementary material, page 4).

In order to select the most promising biomarkers of DCI, we evaluated whether significant biomarkers were validated by a second study. A validated biomarker was defined as a biomarker with significant results in at least two studies using similar methodologies within the same time interval after SAH, without any other study showing contradictory results. Biomarkers of which the meta-analysis of all relevant studies yielded a significant association were considered validated, too. Additionally, biomarkers that had only been investigated by one study were differentiated from those biomarkers that were investigated by a high-quality study. We defined high quality as a maximum (eight or nine) or maximum minus one point (seven out of eight or eight out of nine) quality assessment score.

Data have been reported according to the PRISMA reporting guidelines for systematic reviews and meta-analyses.

Included studies

After removing duplicates, the search yielded a total of 5,820 studies. We excluded 3,900 studies during abstract screening and 1,711 studies during full-text screening, after which 209 articles (204 cohort and 4 case-control studies, 1 randomized controlled trial) remained (Fig. 1). Baseline characteristics of all included studies are listed in Table S1 of the supplemental data (Supplementary material, page 5-12). Altogether, the 209 studies measured 724 different biomarkers in 25,933 SAH patients (not corrected for overlapping patient cohorts). A summary of the NOS quality scores is listed in Table II, and the quality scores of individual studies are displayed in the baseline characteristics table (Table S1, page 5-12). The four case-control studies scored three, four (n = 2) and seven of nine points. The randomized controlled trial scored low risk on three of seven items.

Clinical DCI

Biomarkers of clinical DCI were studied by 166 studies comprising 646 different biomarkers (Fig. 2) in 20,983 SAH patients of whom 5,810 (28%) developed DCI. Most biomarkers (588 of 646, 91%) were investigated by one study and 58 (9%) by more than one study (range 2 to 14, median 3). The distribution of the biomarkers among the biomarker groups is shown in Table III.

In total, 141 biomarkers (n = 88 nongenetic, n = 53 genetic) were classified as significantly associated with clinical DCI (Table S2, page 13-34). A summary of studied and significant biomarkers and specific trends observed within each biomarker group are displayed in Table III. Of the significant biomarkers, 121 were investigated by one study and 20 by more than one study. Meta-analyses were possible for five biomarkers (haptoglobin 1-1, 2-1, and 2-2, ADAM metallopeptidase with thrombospondin type 1 motif 13 [ADAMTS13] and von Willebrand factor [vWF]). The meta-analyses showed that SAH patients with haptoglobin polymorphism 2-1 were less likely to develop DCI and patients with haptoglobin polymorphism 2-2 were more likely to develop DCI (8,9,10). DCI patients had lower ADAMTS13 levels on day 1 and higher levels of vWF on day 4 after SAH, compared to patients without DCI, in two studies (11,12). The results of the remaining 53 significant biomarkers that were investigated by multiple studies that could not be meta-analyzed were summarized in biomarker-specific figures. Based on the biomarker-specific figures, three other biomarkers (neutrophil/lymphocyte ratio [NLR] and P-selectin in blood and F2-isoprostane in urine) had significant results in two or more studies using similar methodologies and timing of measurement, and were considered validated biomarkers. Four studies on NLR showed a proinflammatory state on admission and preoperatively in patients with DCI compared to patients without DCI (13,14,15,16). P-selectin, a marker of endothelial activation, was found to have a significantly larger increase during DCI onset, compared to paired samples in the non-DCI group in two studies (17,18). F2-isoprostane in urine on day 3 after SAH was significantly associated with the occurrence of DCI in two studies (19,20). The quality of studies from which we found the seven validated biomarkers is listed in Table IV. Both meta-analyses and biomarker-specific figures can be found in the supplemental data (Supplementary material, Figure S1-S58, page 35-45).

Among the 121 significant biomarkers investigated by one study only, 5 (n = 4 nongenetic, n = 1 genetic) biomarkers were investigated by high-quality studies: brain natriuretic peptide (21), osteopontin (22), tenascin C (23), and neutrophil/albumin ratio (24) in blood and endothelial nitric oxide synthase (SNP rs2070744 T-786C) using buccal swabs (25).

Radiological DCI

Biomarkers of radiological DCI were studied by 78 studies comprising 165 different biomarkers (Fig. 2) in 10,282 SAH patients of whom 2,254 (22%) developed DCI. Most biomarkers (146 of 165, 88%) were investigated by one study and 19 (12%) by more than one study (range 2 to 8, median 2). The distribution of the biomarkers among the biomarker groups is shown in Table V.

In total, 63 biomarkers (n = 39 nongenetic, n = 24 genetic) were classified as significantly associated with radiological DCI (Supplementary material, Table S2, page 13-34). A summary of the studied and significant biomarkers and specific trends observed within each biomarker group is given in Table V. Of the significant biomarkers, 56 were investigated by one study and 7 by multiple studies. Meta-analyses were not possible. Based on the biomarker-specific figures, one biomarker, lactate/pyruvate ratio (LPR) in blood, was considered validated as two independent studies showed that an LPR higher than 40 was associated with DCI, measured during the first weeks following SAH (26,27). The quality of the two studies on LPR was 4/8 and 2/8 (Table 4). The biomarker-specific figures can be found in the supplemental data (Supplementary material, Figure S59-S77, page 46-48).

Among the 56 significant biomarkers investigated by one study, 7 nongenetic blood biomarkers were investigated by a high-quality study: monocyte LFA-1, monocyte Mac-1, monocyte PSGL-1, neutrophil LFA-1, neutrophil Mac-1, neutrophil PSGL-1, and osteopontin (22,28).

Clinical vs. Radiological DCI

Of all biomarkers, 89 were studied for both clinical and radiological DCI (Supplementary material, Table S2, page 13-34). Eight blood biomarkers were significantly associated with both clinical and radiological DCI (d-dimer, platelet activating factor, interleukin-17, lectin complement pathway activity, osteopontin, brain natriuretic peptide, urea/creatinine ratio and serine proteinase inhibitor family E member 1 SNP rs2227631 + rs1799889 + rs6092 + rs6090 + rs2227684 + rs7242). Forty-five biomarkers were classified as nonsignificant for both clinical and radiological DCI. Of the remaining 36 biomarkers, the classification into significant, inconclusive, or nonsignificant differed between clinical and radiological DCI.