Our search strategy resulted in the retrieval of 3518 unique studies. After careful screening and assessment, nineteen articles were included in the meta-analysis (Supplementary Fig. 1) [11,12,13,14, 25,26,27,28,29,30,31,32,33,34,35,36,37,38,39]. Eight studies were classified as RCTs [14, 25, 29, 30, 33, 37,38,39], six as retrospective cohort studies [11, 12, 28, 31, 34, 35], and five as matched case–control studies [13, 26, 27, 32, 36]. The majority of studies focused on IVH in the setting of IPH (68%) compared to SAH (26%), with the remainder including both IPH and SAH patients (5%). The use of recombinant tissue plasminogen activator ((r)t-PA) increased over the years and was slightly more common than urokinase, 55% versus 45% of the studies, respectively. Additional study details can be found in Table 1.
Collectively, 1020 patients were included in this meta-analysis of which 526 received intraventricular fibrinolytics. Male patients constituted 55.9% of the studied population, and the overall mean age was 56 years (median 56 years). Patients who received IVF had a mean age of 56.0 years (median 55.5), while those receiving EVD only had a mean age of 56.1 years (median 56). More study details and the assessment of bias regarding RCTs can be found in Table I, Supplementary Tables 2–5, and in Supplementary Results section.
Eighteen studies reported on mortality in patients with IVH [11,12,13,14, 25,26,27,28,29,30,31, 33,34,35,36,37,38,39]. Pooled analysis showed a significant decrease in mortality risk for patients receiving IVF treatment with EVD compared to patients receiving EVD alone in both the fixed- and random-effect models, with RR 0.58 (95% CI 0.47–0.72) for both models (Fig. 1). Heterogeneity was low in the random-effects (RE) model (I2 = 0%, p-heterogeneity = 0.89). Meta-regression found no sources for confounding. Egger’s test and Begg’s test were not significant, with p = 0.14 and p = 0.15, respectively. The funnel plot showed a possible bias but correction via the trim-fill method (Supplementary Fig. 2) did not yield a significantly different model.
Functional outcome was assessed in eight studies totaling 749 patients [12, 14, 28, 29, 32, 33, 36, 38]. GFO did not differ after IVF treatment compared to patients receiving EVD alone with RR 1.41 (95% CI 0.98–2.03), in both fixed- and random-effect models (Fig. 2). Heterogeneity of the RE model was moderate, at 32.9%, p = 0.11. Meta-regression analysis showed that study quality was an effect modifier with β = 0.83, p = 0.005, as was study design, with retrospective cohort studies having a β = 0.90, p = 0.006, with RCT as reference category. Curiously, the impact factor of the journal in which the study was published was also an effect modifier, with β = 0.02, p = 0.03, as was the country in which the study was done: studies in Europe modified the effect with β = 0.52, p = 0.03, compared to North-American studies (Supplementary Table 6). Egger’s test was significant at p = 0.04, Begg’s test was not, p = 0.11. The funnel plot showed possible bias in reporting, but correction via the trim-and-fill method (Supplementary Fig. 3) did not predict a significant new model, with RR 1.31 (95% CI 0.93–1.85).
Information regarding the incidence of ventriculitis was available from 15 studies [12,13,14, 25,26,27,28,29,30,31, 33,34,35, 37, 38]. Ventriculitis rates were not significantly lower in patients using IVF with EVD compared to patients receiving EVD alone: 0.68 (95% CI 0.45–1.03) for both the fixed- and random-effect models, p = 0.06 (Supplementary Fig. 4). There was no evidence for heterogeneity in the RE model (I2 = 0%, p = 0.97), and no factor was identified as a source of heterogeneity through meta-regression. Egger’s test (p = 0.44) and Begg’s test (p = 0.86) were not significant. The funnel plot showed a possible indication for bias (Supplementary Fig. 5), and correction via the trim-fill method did yield a significant model, with a decreased risk of ventriculitis (RR 0.61, 95% CI 0.41–0.91, p = 0.02), with no heterogeneity (0%, p = 0.94).
Symptomatic intracranial bleeding after start of therapy was evaluated in 926 patients in 14 studies [12,13,14, 25,26,27,28, 30, 31, 33,34,35,36, 38, 39]. The fixed- and random-effect models showed no significant impact of the treatment on outcome (RR 1.50, 95% CI 0.89–2.52), with low heterogeneity (I2: 0%, p = 0.99) in the RE model (Supplementary Fig. 6). Meta-regression showed that none of the factors contributed to heterogeneity. Egger’s (p = 0.44) and Begg’s (p = 0.85) tests were not significant, but the funnel plot showed possible bias (Supplementary Fig. 7). Correction for publication bias via the trim-fill method predicted a significant random-effects model with RR 1.67 (95% CI 1.01–2.74) with low heterogeneity (I2: 0%, p = 0.99).
EVD obstruction rate was evaluated in seven studies totaling 185 patients [26,27,28, 31, 33, 34, 36]. Obstruction rates were significantly lower in patients treated with IVF compared to patients receiving EVD alone with a RR of 0.41 (95% CI 0.22–0.74) in both the fixed- and random-effect models (Fig. 3). In the RE model, heterogeneity was low at I2 = 0% with p = 0.79. No sources of heterogeneity were identified by meta-regression. Funnel plot, Egger’s test (p = 0.26), and Begg’s test (p = 1.0) did not indicate publication bias.
Time to IVH Resolution
Clearance of the third and fourth ventricles was assessed in six studies totaling 596 patients [14, 26, 27, 30, 31, 38]. The random-effects model showed a significantly faster ventricular clearance in patients receiving IVF with EVD compared to patients receiving EVD alone (mean difference − 4.05 days, 95% CI between − 5.52 and − 2.57) (Fig. 4). The fixed-effect model showed similar, yet slightly weaker, results (mean difference − 3.27 days (95% CI between − 3.57 and − 2.97). Heterogeneity was high in the RE model, however, with I2: 91.3%, p < 0.0001. Study country (Europe: β = −3.82, p < 0.001, Middle East: β = −1.29, p = 0.0001, North America as reference), design (case–control: β = −2.81, p = 0.035, RCT as reference), size (β = 0.003, p < 0.0001), and the impact factor of the journal in which the study was published (β = 0.03, p < 0.0001) all interfered with outcome, as did patient age (β = 0.46, p < 0.0001) (Supplementary Table 6). The funnel plot, Egger’s test (0.52), and Begg’s test (0.72) indicated low possibility of bias.
Shunt dependency after IVH was assessed in 16 studies [12,13,14, 26, 27, 29,30,31,32,33,34,35,36, 38, 39]. There was no difference between IVF + EVD and EVD alone in the risk shunt dependency using the random-effect or fixed-effect models (RR 0.93, 95% CI 0.70–1.22, p = 0.59). No heterogeneity was found in the RE model, I2: 0%, p = 0.71. Average age was identified as a possible source of heterogeneity, with β = 0.11 (p = 0.048) (Supplementary Table 6). Egger’s and Begg’s tests were not significant, with p = 0.27 and p = 0.33, respectively. The funnel plot showed possible publication bias (Supplementary Fig. 8), but correction via the trim-fill method did not yield a significantly different model.
IVF in SAH Versus IPH
Since SAH and IPH have very different underlying pathologies, we analyzed differences between these patient groups. The origin of the IVH (IPH or SAH) showed no significant impact on any of the outcomes in our meta-regression analysis (Supplementary Table 6). We performed a subgroup analysis to further evaluate this finding, comparing IPH only studies (12 studies, 885 patients) with SAH only studies (five studies, 109 patients). Risk of mortality, obstruction, and clearance of the ventricles remained significantly improved in IPH patients receiving IVF, while only the risk of faster clearance of the ventricles was significantly improved in SAH patients with IVF (data not shown).