Over the past decade, there has been intensive research into enhancing the effects of reperfusion therapy by mitigating its adverse consequences such as reperfusion injury and hemorrhagic transformation (HT), and promoting functional recovery in stroke patients (Saver, 2011). The most straightforward and promising approach appears to be the use of neuroprotective agents with multimodal pleiotropic properties along with intravenous thrombolysis (IVT) (Otsu et al., 2020). Cerebrolysin, a porcine brain derivate, is one of the candidates since it consists of low-molecular weight neuropeptides and free amino acids, which mimic the action of endogenous neurotrophic factors on protection and repair of the central nervous system (Masliah and Díez-Tejedor, 2012).

Since early 1970-s, Cerebrolysin has been widely used in treatment of acute ischemic stroke, traumatic brain injury and cognitive impairment in over 50 European, Middle East, Latin American and Asian countries but not in the United States, United Kingdom and Australia (Cui et al., 2019; Staszewski et al., 2022; Jarosz et al., 2023; Ziganshina et al., 2023). A plethora of clinical trials and meta-analyses have suggested Cerebrolysin enhances early post-stroke recovery, improves neurological deficit after stroke, and bears an excellent tolerability and safety profile (Lang et al., 2013; Guekht et al., 2017; Bornstein et al., 2018). Therefore, it has been included in the guideline on pharmacological support in motor rehabilitation after acute ischemic stroke issued by the European Academy of Neurology and the European Federation of Neurorehabilitation Societies (Beghi et al., 2021).

In our original study published in March 2023 (Khasanova and Kalinin, 2023), we looked at the effects of Cerebrolysin with IVT versus IVT alone in stroke patients. The rationale behind the trial came from two aspects. First, recombinant tissue plasminogen activator (rtPA) increases the HT rate by degrading the blood-brain barrier (BBB) integrity and promoting neuroinflammation and excitotoxicity (Wang et al., 2004; Kelly et al., 2006; Guan et al., 2019). On the other hand, Cerebrolysin ameliorates rtPA adverse effects and, therefore, can potentially protect from IVT-related HT (Veinbergs et al., 2000; Teng et al., 2021; Guan et al., 2019). Besides bearing neuroprotective properties directed to mitigate the reperfusion injury, Cerebrolysin also attenuates BBB permeability (Zhang et al., 2010; Teng et al., 2021), which plays an essential role in HT pathophysiology (Spronk et al., 2021).

Thus, Cerebrolysin as an early add-on to IVT turned out to be beneficial for stroke patients: the treatment was safe, and reduced the symptomatic HT rate and early neurological deficit. Furthermore, our findings were coherent with the results of a similar trial on patients with severe stroke and futile recanalization after IVT (Poljakovic et al., 2021).

However, subjects differ not only in their background characteristics, but also in how they respond to a particular treatment or intervention (Xie et al., 2012). Accordingly, it is reasonable to assume that patients’ response to the Cerebrolysin treatment would vary with respect to their HT risk stratification on admission, and hence, discovering a group of patients benefiting most by the Cerebrolysin treatment could be of paramount importance for decision making in acute stroke settings. Therefore, we conducted post hoc analysis of the CEREHETIS trial (Khasanova and Kalinin, 2023) to look at treatment effects of Cerebrolysin as an early add-on to the reperfusion therapy in stroke patients with varying HT risk.

CEREHETIS was a prospective, randomized, open-label, active control, multicenter, parallel-group phase IIIb pilot study (trial registration number: ISRCTN87656744). The trial protocol, patient inclusion/exclusion criteria and original results have been published previously (Khasanova and Kalinin, 2023).

Each eligible patient was randomly assigned into either the Cerebrolysin or control group. Both arms received a standard dose of 0.9 mg/kg alteplase (Actilyse^®, Boehringer Ingelheim GmbH, Germany), rtPA, administered intravenously within 4.5 h after symptom onset (maximal dosage 90 mg, 10% of the drug given in bolus and the rest in 60 min via intravenous infusion). Patients in the intervention group were additionally given 30 mL of Cerebrolysin^® (EVER Pharma GmbH, Austria) diluted in 100 mL of normal saline and infused intravenously through a separate line over 20 min. The Cerebrolysin treatment was initiated simultaneously with IVT and continued once daily for 14 consecutive days.

The study primary endpoints were the rate of any and symptomatic HT verified on a follow-up computed tomography scan from day 0 to day 14. Symptomatic HT was defined according to the ECASS III trial: any apparently extravascular blood in the brain or within the cranium that was associated with clinical deterioration, as defined by an increase of 4 points or more in the score on the National Institutes of Health Stroke Scale (NIHSS), or that led to death and that was identified as the predominant cause of the neurologic deterioration (Hacke et al., 2008).

The secondary endpoint was functional outcome measured with the modified Rankin Scale (mRS) on day 90. Favorable functional outcome (FFO) was defined as an mRS score of ≤2 on day 90.

Of 341 participants comprising the intention-to-treat population of the CEREHETIS trial, 238 MCA stroke patients were selected for analysis (Figure 1).

At baseline, patients in the Cerebrolysin arm were slightly younger and, as a result, had fewer cases of previous stroke (Table 1). That was of no surprise, since the covariate disbalance had already been observed in the original cohort (Khasanova and Kalinin, 2023).

Although the score distribution did not extend over the full scale for the DRAGON (0–10), SEDAN (0–6) and HTI (0–8) score (Table 1) due to the study exclusion criteria (e.g., patients with neither minor (NIHSS <4) nor severe (NIHSS >25) stroke nor high hemorrhagic risk were eligible for IVT), each tool was able to predict HT and FFO and confounded the Cerebrolysin treatment (Figures 2A–C). However, only the HTI score became significant in the combined model (Figure 2D) and its performance was superior to the competitors in the postestimation tests (Table 2), hence it was selected for further analysis.

The summarized output of the employed statistical techniques–two-way contingency tables, logistic regression and propensity score matching followed by the treatment effect assessment–was consistent with our original results (Khasanova and Kalinin, 2023): the Cerebrolysin treatment significantly reduced the symptomatic HT rate and outlined a tendency to alleviate any HT and poor functional outcome (Table 3; Figures 2C, 3A, C).

Since CEREHETIS was a prospective randomized trial with almost equal covariate distribution between the treated and the untreated (Table 1), the estimated treatment effects (ATE, ATT, ATC, NATE) and their corresponding potential outcome averages did not differ much from one another (Figure 3C), and the HTI score had no influence on the treatment assignment (propensity score matching: HTI coefficient = 0.006; 95% CI: −0.258, 0.270; p = 0.962).

Based on the HT predicted probability adjusted for the HTI score (Table 4), HT risk can be graded for the sake of simplicity as low (HTI = 0), moderate (HTI = 1) and high (HTI ≥2).

In that respect, the assessment of conditional marginal effects of the Cerebrolysin treatment on probability of symptomatic HT by the HTI score revealed a clear pattern–the impact was neutral in the low HT risk patients but became prominent in those with the moderate and high risk. Accordingly, the Cerebrolysin treatment decreased the symptomatic HT rate by 2.4% (p > 0.05) in patients with HTI = 0, which further dropped step-wise to 32.5% (p < 0.05) at HTI = 4 (Figure 3B).

The mRS stacked bar plot arranged by the HTI score demonstrated two positive trends in functional outcome of the Cerebrolysin group. First, at HTI = 0 and 1, there was a decrease in the percentage of patients with mRS = 3 on day 90 (HTI = 0: 0% in the Cerebrolysin arm vs. 9% in the control; HTI = 1: 15% vs. 30%, respectively). Second, the overall percentage of patients with FFO was higher in the Cerebrolysin group. Interestingly, it was more prominent at HTI = 3 (50% vs. 9%, respectively) (Figure 4).

The Brant test turned out to be insignificant (χ² (10) = 4.31, p = 0.932), hence the parallel regression assumption was not violated and generalized ordered logistic regression with proportional as well as partial proportional odds was justified for analysis. Both approaches defined the HTI score as a significant predictor of functional outcome (Figure 5).

After applying the partial proportional odds assumption, the aforementioned first trend did achieve the statistical significance but was concealed by the use of a less flexible technique, the parallel one. However, both procedures failed to provide robust estimates of the second trend (Figures 5, 6).

Although the applied statistical approaches did establish some clinical variability in Cerebrolysin treatment effects among patients with varying HT risk, their heterogeneity was confirmed using methods of meta-analysis. The subgroup of patients with HTI = 4 was omitted during the procedure due to missing values in the Cerebrolysin arm (Table 3).

The overall and subgroup direction and size of Cerebrolysin treatment effects on HT and functional outcome did not differ much between the general population (the random-effects model) and studied cohort (the fixed-effects model) and were coherent with the results of the previous tests. By the way, no small-study effects, the phenomenon that smaller subgroups (“studies”) showed different treatment effects than large ones, were found in any model of interest (Egger test, p > 0.05). Heterogeneity of Cerebrolysin treatment effects turned out to be moderate (I², 35.8%–56.7%; H², 1.56–2.31) and mild (I², 10.9%; H ² , 1.12) for symptomatic and any HT, respectively. However, the approach was unrevealing for functional outcome (Figures 7, 8).

Nevertheless, HTE analysis by means of the matching-smoothing method demonstrated a significant positive impact of the Cerebrolysin treatment on HT and functional outcome: it was neutral when the risk of HT and poor functional outcome was low (HTI = 0) but gradually became evident in the moderate (HTI = 1) and high (HTI ≥2) HT risk patients. Largely, Cerebrolysin HTEs appeared in a linear-slope manner (Figure 9).

In particular, there was a steady decline in the rate of symptomatic (HTI = 0 vs. HTI = 4: by 4.3%, p = 0.077 vs. 21.1%, p < 0.001, respectively) and any HT (HTI = 0 vs. HTI = 4: by 1.2%, p = 0.737 vs. 32.7%, p < 0.001, respectively). Likewise, Cerebrolysin treatment resulted in an overall reduction of the mRS score, which was significantly greater with increasing HTI scores (HTI = 0 vs. HTI = 4: by 1.8%, p = 0.903 vs. 126%, p < 0.001, respectively). Reciprocally, a growing fraction of patients with FFO (HTI = 0 vs. HTI = 4: by 1.2% p = 0.757 vs. 35.5%, p < 0.001, respectively) was observed with climbing HTI scores (Table 5; Figure 9).

Thus, eligible for IVT patients with estimated on-admission HT risk of the moderate or high grade turned out to be the most responsive to the Cerebrolysin treatment.

Using sophisticated statistical analysis, our post hoc study established clinically meaningful heterogeneity of Cerebrolysin treatment effects with respect to HT mitigation and functional outcome improvement among patients with varying HT risk. More importantly, the positive impact of the Cerebrolysin treatment grew steadily along with increasing HT risk. In general, the study results were coherent with the original report (Khasanova and Kalinin, 2023) and discovered a group of patients benefiting most by the proposed treatment.

Treatment effect refers to the causal effect of a treatment or intervention on an outcome of interest based on the counterfactuals (e.g., difference in outcomes with/without using the drug). HTE analysis focuses on assessing varying treatment effects for individuals or subgroups in a population (Gong et al., 2021).

Estimation of HTEs plays an essential role in randomized clinical trials, in particular, to identify patients who benefit most by the intervention for developing personalized treatment (Imai and Ratkovic, 2013; Ling et al., 2023). Meanwhile, traditional statistical approaches like descriptive statistics and regression analysis have a strong tendency to focus on the significance of the estimated overall ATE rather than systematically evaluate the variation in treatment effects across subgroups. As a result, there has been the standard practice to report a number of single estimates representing the treatment efficacy (Imai and Strauss, 2011).

HTEs are usually examined with a predictive model for individual outcomes followed by the exploration of interactions between treatment allocation and important patients’ baseline characteristics. In contrast, our HTE analysis was in favor of the two-stage approach (Watson and Holmes, 2020) since it allowed us to avoid overfitting and transformations of the outcome measurements.

Moreover, assessment of HTEs could especially be useful in settings in which univariate subgroup analyses are unrevealing (Raghavan et al., 2022). For instance, heterogeneity of treatment effects alone may be considered as a sufficient explanation for negative results in studies of some neuroprotective agents because even in the largest trials sample size is inadequate to detect effect size equivalent to those with IVT (Samsa and Matchar, 2001; Muir, 2002). Indeed, various regression approaches failed to provide robust estimates of the functional outcome improvement in our study, whereas HTE analysis did unambiguously confirm those patterns.

However, HTE analysis has not been widely used despite the fact that treatment effects are rarely perfectly homogeneous over the population. Perhaps, part of the reason is the complexity of its methods (Gong et al., 2021). Indeed, only few studies addressing HTEs in stroke patients have been published by now (Kent et al., 2001; Kent et al., 2021; Zhou et al., 2022; Ngufor et al., 2023; Zhu et al., 2023), but none of them related to neuroprotection. Therefore, we have striven to adhere meticulously to the proposed guidelines (Gabler et al., 2009) to report the results of our HTE analysis.

A number of tools have been suggested to predict HT in stroke patients (Kalinin et al., 2017). Although many of them are reliable and composed of the same set of predictors with some variations, only the HTI score takes into account a vascular territory of the infarcted brain lesions. This could be a reason why the HTI score outperformed the DRAGON and SEDAN scores in HT risk stratification in MCA stroke patients.

There are numerous well-established differences between stroke in the posterior and anterior circulation (Libman et al., 2001; Sarikaya et al., 2011; Merwick and Werring, 2014), which could be a source of additional heterogeneity of treatment effects. Moreover, some HT risk assessment tools turned out to be of little value in patients with posterior circulation stroke (Sung et al., 2013). Therefore, only MCA stroke patients were included in the study because we considered them to a certain extent as a homogeneous cohort from the anatomical point of view.

Permeability–surface area product (PS) is a known perfusion imaging marker of BBB permeability and an independent HT predictor. The HTI score is able to predict not only the HT probability but also brain perfusion data, in particular, PS values. The more PS rises in the infarct core following stroke, the higher the HTI score, and hence, the higher the HT risk (Kalinin et al., 2019).

Moreover, rtPA increases the HT rate by degrading the BBB integrity and promoting neuroinflammation and excitotoxicity (Wang et al., 2004; Kelly et al., 2006; Guan et al., 2019). On the other hand, Cerebrolysin ameliorates rtPA adverse effects (Veinbergs et al., 2000; Teng et al., 2021; Guan et al., 2019), and our original report has confirmed neuroprotective and BBB stabilizing features of Cerebrolysin in clinical settings: an improvement in the diffusion-tensor imaging metrics and PS was observed in the infarcted lesions (Khasanova and Kalinin, 2023). Therefore, benefiting most by the Cerebrolysin treatment in HT-risky circumstances could be explained by more efficient salvage of the ischemic brain tissue as well as more prominent attenuation of BBB permeability and mitigation of rtPA adverse effects with the Cerebrolysin use.

Based on high-quality evidence, current clinical guidelines strongly recommend IVT for patients with acute ischemic stroke of <4.5 h duration (Berge et al., 2021). Since the earliest publications of systematic reviews and meta-analyses on IVT-related HT, a paradigm that no reliable method can provide the individualization of treatment according to predicted HT risk has dominated (Whiteley et al., 2012), which practically makes all available prediction tools useless. As a result, prior HT risk assessment is not even required to start IVT in stroke patients. In that respect, our research is one of the first steps to shift that paradigm.

Several clinical studies have investigated the concomitant use of various agents, which reduce the HT rate and exert multimodal effects, alongside IVT. However, phase III clinical trials are required to confirm the observed positive results (Otsu et al., 2020). The ESCAPE-NA1 trial, a large-scale study of the neuroprotective agent nerinetide, failed to demonstrate an improvement in functional outcome due to a possible drug-drug interaction with alteplase (Hill et al., 2020). In that respect, Cerebrolysin could be an alternative to nerinetide and other candidates given its ability to mitigate rtPA-related adverse effects, well-established safety profile and long-lasting use in clinical practice worldwide.

Clinical practitioners frequently encounter so called the risk-treatment paradox, in which patients at the highest risk (and with the greatest potential to gain from the treatment) are treated less often than those at lower risk and with less potential to benefit (Spertus et al., 2015). In that respect, our study is another step towards integrating individualized risk stratification within routine clinical practice to disregard this misconception, to remind clinicians of each patient’s potential benefits from the treatment and to ensure more patient-orientated, evidence-based care with favorable outcomes.

The strength of the study emerged from stratification of stroke patients according to their HT risk followed by a complex statistical approach to analyze heterogeneity of Cerebrolysin treatment effects.

Nevertheless, there are a few limitations in our study. First, an anatomical restriction: only MCA stroke patients were selected for analysis. Heterogeneity of treatment effects in subjects with vertebrobasilar infarction can significantly affect the outcome of interest and requires a separate subanalysis.

Second, an incomplete range of the HTI score: due to study exclusion criteria, the population with an HTI score of ≤4 was only analyzed. Obviously, patients with an HTI score beyond this limit would fall into a category of extremely high risk of HT and would have a large infarct core and severe stroke. In those circumstances, IVT might be considered in some selected cases based on the results of advanced brain imaging (core/perfusion mismatch) and other contraindications (Berge et al., 2021). In that respect, a clinical trial on patients with severe stroke and futile recanalization after IVT has suggested Cerebrolysin as an add-on to IVT is safe and reduces the HT and mortality rate (Poljakovic et al., 2021). Moreover, recent clinical trials on mechanical endovascular thrombectomy (EVT) in patients with a large infarct core (Yoshimura et al., 2022; Huo et al., 2023; Sarraj et al., 2023) have shown a significant improvement in functional outcome compared with the standard medical care alone despite an increased HT rate in the intervention group. Thus, it is fairly reasonable to assume the Cerebrolysin treatment combined with reperfusion therapy could result in a reduction of the HT rate and further enhancement of functional outcome in those patients.

At last, post hoc analyses inherently suffer from a well-known statistical problem: the subgroups are formed after the trial is conducted, and such analyses bear a risk of finding statistically significant results when no true relationship exists (Imai and Strauss, 2011). In that respect, there are a number of challenges related to researcher bias in secondary data analysis such as prior knowledge of data, non-hypothesis-driven research, inappropriateness and lack of flexibility in data analysis (Baldwin et al., 2022). Therefore, a variety of approaches were applied to address this issue (Table 6). Yet, further research with prespecified HT risk subgroups is required.

Clinically meaningful HTEs of Cerebrolysin as an early add-on to reperfusion therapy were established in patients with varying HT risk. In terms of FFO achievement and HT rate reduction after IVT, the Cerebrolysin treatment appeared to be beneficial in those whose estimated on-admission HT risk was either moderate or high. The evidence is sufficient to guide different treatment recommendations in one or more subgroups but warrants future research with prespecified hypotheses.