Caffeine as a Treatment for Perinatal Hypoxic-Ischemic Brain Injury: The Potential Risks and Benefits

Moderate-to-severe hypoxic-ischemic encephalopathy (HIE) affects around 1–3/1,000 live term births in the developed world and is associated with a high rate of death or lifelong disability, including cerebral palsy [1]. Although therapeutic hypothermia significantly improves survival without disability for term infants with moderate-to-severe HIE into middle childhood in high-income countries, ∼29% of infants treated with therapeutic hypothermia still survive with disability [1‒3]. Furthermore, the hypothermia for moderate or severe neonatal encephalopathy in low-income and middle-income countries (HELIX) trial found that therapeutic hypothermia did not affect the combined risk of death or disability and was associated with an increased risk of death in cooled infants [4]. The concerning findings of this study have drawn into question whether therapeutic hypothermia should be offered in this setting. Given that no other approved treatment strategies are available, there is an urgent need to identify effective treatments for use in low- and middle-income countries, as well as effective add-on treatments with therapeutic hypothermia to further improve outcomes for HIE in high-income countries.

Caffeine is widely used in neonatal intensive care units to prevent or treat apnea in preterm infants, with the long-term goal of improving neurodevelopmental outcomes. Critically, in the largest randomized controlled trial of caffeine for preterm infants – The Caffeine for Apnea of Prematurity (CAP) trial, caffeine was associated with a reduced incidence of cerebral palsy, cognitive delay, hearing loss, and blindness at 2 years of age [5]. In an 11-year follow-up study of the CAP trial, children who had received caffeine treatment had improved visuomotor, visuoperceptual, and visuospatial abilities and reduced motor impairment, albeit with no differences in general intelligence, behavior, and attention, compared with the placebo group [6, 7]. It is unclear whether these protective effects of caffeine are simply secondary to reduced intermittent hypoxia or whether they could be a direct neuroprotective effect.

Given the encouraging evidence in preterm infants, caffeine has been tested as a neuroprotective agent for perinatal hypoxic-ischemic (HI) brain injury in preclinical models. In this narrative review, we discuss the mixed results of preclinical data to date and the very limited clinical data on the safety and efficacy of caffeine for HIE.

Caffeine (1,3,7-trimethylxanthine) is a methylxanthine derivative that acts as an antagonist for all adenosine receptors – A₁, A_2A, A_2B, and A₃ (Fig. 1). At low doses, caffeine competitively inhibits A₁ and A_2A receptors, and at higher doses, it inhibits phosphodiesterase enzymes [8]. At supraphysiological levels, caffeine can lead to calcium release from the endoplasmic reticulum, and at toxic levels, caffeine has low affinity for gamma-aminobutyric acid (GABA_A) receptors as an antagonist [8, 9].

The active component of caffeine citrate is caffeine base, which comprises 50% of the total dose. In this review, we will specify whether caffeine citrate or base is used in a study, unless it was not stated. Caffeine is lipophilic and therefore can cross the blood-brain barrier easily, supporting its potential for neuroactive effects. Of note, the plasma elimination of caffeine is considerably longer in preterm newborns (102 h) than adults (3–6 h) [10]. In adults, there is dose-dependent kinetics of caffeine due to the saturation of pathways involving CYP1A2 and CYP2E1 [11]. In newborns, there were concerns that high doses of caffeine would exhibit nonlinear kinetics, which would decrease elimination and increase the risk of toxicity. Reassuringly, maintenance doses of 3, 15, 30 mg/kg/day for 7 days in newborns did not show dose dependent kinetics [12].

Adenosine is an endogenous purine nucleoside that acts as a ubiquitous signaling molecule with pleiotropic effects that are fundamental in many organs of the body, including in the brain. Adenosine is often referred to as a reactive metabolite, as extracellular levels of adenosine significantly increase during HI due to the hydrolysis of ATP to ADP and AMP, and then to adenosine. Adenosine is also released from cells via the equilibrative nucleoside transporter 2 [8, 13]. After hypoxia or increased metabolic activity, levels of adenosine in the brain can increase over 100-fold in the extracellular space [14, 15].

Adenosine receptors or P1 receptors are G-protein-coupled receptors, and there are four known types of adenosine receptors – A₁, A_2A, A_2B, and A₃ [8]. A₁ and A_2A receptors are high-affinity receptors, whereas A_2B and A₃ are low-affinity receptors [8, 16, 17]. In the brain, adenosine receptors are expressed on neurons, glia, and immune cells [8]. The wide distribution of adenosine receptors highlights that adenosine is an important neuromodulator that has widespread effects in the brain.

In this review, we will focus on the effects on A₁ and A_2A receptors as these two receptors have been most well-studied in the context of perinatal HI. It is widely accepted that in the adult brain, A₁ receptor activation is protective against HI injury and that these protective effects are blocked by A₁ receptor antagonists [8, 18, 19]. Surprisingly, in the developing brain, there is highly conflicting evidence on the roles of each of the adenosine receptors and whether agonism or antagonism of these receptors is neuroprotective.

Adenosine is a key factor in the rapid suppression of brain activity at the onset of profound HI. When a highly specific adenosine A₁ receptor antagonist 8-cyclopentyl-1,3-dipropylxanthine (DPCPX) was infused before 10 min of umbilical cord occlusion in near-term fetal sheep (equivalent to the human brain at term), there was a transient increase in electroencephalogram activity followed by a delayed fall, compared with the rapid and sustained suppression seen in controls [20]. Further, DPCPX was associated with earlier onset of cytotoxic edema during asphyxia and a smaller reduction in heat production. After reperfusion, there was a greater seizure burden and neuronal loss compared with controls [20]. These data suggest that blocking adenosine A₁ receptor prevents the endogenous neuroprotective effects of adenosine during profound hypoxemia, including its ability to rapidly suppress brain activity and metabolism, and consequently delay the onset of brain injury. Further, there is also evidence that adenosine and A₁ receptors play an essential role in protecting the embryo from hypoxia. For example, homozygous A₁ receptor knockout mice embryos had more severe growth restriction when compared to wild-type or heterozygous knockouts [21]. In contrast, administration of DPCPX to P7 rats after to common carotid artery ligation and hypoxia had no effect on brain injury [22].

Interestingly, A₁ receptor agonism is also reported to have inconsistent effects. Treatment with the A₁ receptor agonist adenosine amine congener was not neuroprotective when given to postnatal day 7 (P7) rats after common carotid artery ligation and hypoxia [23]. This lack of effect may suggest that the A₁ receptor is not well developed yet in P7 rat pups (human late-preterm equivalent). Against this hypothesis, under normoxic conditions in normal neonatal rats, prolonged, high-dose treatment with A₁ adenosine receptor agonist N6-cyclopentyladenosine from P3 to P14 was associated with reduced white matter volume, ventriculomegaly, and neuronal loss [24]. This suggests that these receptors are indeed functional at early postnatal age in the rat but speculatively may respond differently to different agonists.

It is important to appreciate that adenosine plays a key role in terminating seizures; this antiseizure effect has been known for decades [25]. Extracellular levels of adenosine are increased by local tissue hypoxia during intense metabolic activity, such as seizures, and mediate postictal EEG suppression [25‒27]. Elevated levels of extracellular adenosine in the range of 25–250 nM inhibit neuronal activity via the inhibition of presynaptic neurotransmitter release [28]. The effects of adenosine on seizure control appear to be predominantly mediated by the A₁ receptor, as blocking of the receptor using DPCPX increased the susceptibility to status epilepticus in rats after repeated electrical stimulation [29]. Conversely, administration of the selective A₁ receptor agonists N6-cyclohexyladenosine and N6-cyclopentyladenosine prevented the development of status epilepticus [29].

The evidence for a role of the A_2A receptors after HI in the developing brain is limited and mixed. In P7 mice exposed to common carotid artery ligation and hypoxia, A_2A selective antagonism with either SCH 58261 or the nonselective adenosine receptor antagonist theophylline reduced brain injury [22]. By contrast, A₂ adenosine receptor knockout mice that were subjected to common carotid ligation and hypoxia at P7 had more severe brain injury with worse performance in motor behavioral tests, compared with wild-type mice [30].

In most preclinical studies on the effect of caffeine for perinatal HI brain injury, single or repeated doses of caffeine were often given before or immediately after HI (Table 1). In P7 rats subjected to common carotid artery ligation and hypoxia, caffeine citrate (20 mg/kg/day, intraperitoneally, i.p.) given immediately before HI and at 0, 24, 48, and 72 h after HI was associated with reduced cell loss in the hippocampus and cortex, compared with the saline group [31]. In preterm equivalent (P3) rats exposed to common carotid artery ligation and hypoxia, caffeine citrate 20 mg/kg/day i.p. administered from day 2–6 [32] was associated with reduced ventricle dilation and improved myelin expression in the caffeine group, compared to the vehicle group. Furthermore, caffeine was associated with the inhibition of NOD-, LRR-, and pyrin domain-containing protein 3 (NLRP3) inflammasome activation, reduced microglial activation and microglia M1 polarization, and promoted microglia M2 polarization [32]. Interestingly, in this study, administration of an A_2A receptor agonist (CGS21680) reversed the effects of caffeine [32].

Further, there is evidence in neonatal rodents for protection with very early treatment after HI. A single dose of caffeine (5 mg/kg i.p.) given immediately after common carotid artery ligation and hypoxia reduced gray and white matter lesion size, reduced the number of amoeboid microglia and apoptotic cells in P10 mice, compared with the phosphate buffered saline group [33]. Similarly, a single dose of caffeine (5 mg/kg) in the same model resulted in reduced brain atrophy and increased time on the rotorod behavioral test [34]. When caffeine (10 mg/kg i.p.) was administered to P7 male rats immediately after common carotid artery ligation and hypoxia, this resulted in partially increased cortical volume and improved performance in the Morris water maze [37]. In P6 rat pups subjected to common carotid artery ligation and hypoxia, caffeine citrate (20 mg/kg i.p.) administration immediately after HI was associated with better performance in the rotarod and water maze behavioral test and improved silent gap detection (speech detection) compared with saline treated animals assessed at P35 [35]. Despite improved behavioral outcomes, there was only partial preservation of brain volume with caffeine treatment [35]. By contrast with these encouraging results, when administration of caffeine was delayed to 6, 12, or 24 h after HI, there was no neuroprotective effect [33]. Additionally, it should be noted that both of these studies had no sham control group [33, 34].

In a caffeine dose-response study in P7 rats, 40 mg/kg (i.p.) given after carotid artery ligation but before hypoxia with repeated injections at 24 and 48 h was associated with the least brain area loss, compared with the 15 mg/kg or 20 mg/kg groups [36]. When the first injection of caffeine (40 mg/kg i.p.) was given after 90 min of hypoxia, this significantly reduced brain area loss, but there was greater variability in the response when compared with starting caffeine before hypoxia [36]. Bernis et al. [36] showed that there was also reduced microgliosis with caffeine treatment and that the neuroprotective effects were mediated through modulating the AMPK/mTOR pathway. Moreover, a 120 mg/kg dose of caffeine resulted in 100% mortality in P7 rats after the second dose at 24 h [36].

A multidrug screening trial of the most promising neuroprotective agents for neonatal HI in P7 rat pups exposed to common carotid artery ligation and hypoxia [38] found that 8 of the 25 agents tested were associated with a significant reduction in brain area loss. In this trial, caffeine given 1 h before hypoxia and then at 24 and 48 h (40 mg/kg i.p.) had one of the strongest treatment effects. In this study, with drugs that are able to cross the placental barrier, the first dose was given before HI. Interestingly, the probability of efficacy was greater for caffeine treatment than therapeutic hypothermia (100% vs. 98.8%, respectively) [38]. However, this is likely related to the suboptimal duration of cooling of 5 h used in rat pups, compared with recommended protocol of 72 h of cooling clinically. Pragmatically, in a clinical setting, pretreatment is rather unlikely to be feasible. From the perspective of clinical translation, disappointingly, delayed post-HI treatment was not tested.

There are few caffeine studies in large animal translational models. Recently, in lambs subjected to umbilical cord occlusion at day 141–143 gestation (brain development equivalent to a human toddler [40]) [39], both prophylactic and post-insult caffeine treatment was tested. 1 g (intravenously [i.v.]) was given to the ewe before delivery, and then the lambs received 20 mg/kg caffeine citrate after delivery followed by 2 doses of 10 mg/kg i.v. at 24 and 48 h. Another group of lambs received treatment after birth only with an initial high dose of caffeine (60 mg/kg i.v.), followed by 30 mg/kg i.v. at 24 and 48 h. This high-dose group had greater mortality and was excluded from subsequent analysis. It should be noted that the injury induced by HI in this study was very mild, with selective neuronal loss in the CA1 of the hippocampus and an increase in astrocytes and microglia. Prophylactic low-dose caffeine treatment did not improve hippocampal neuronal survival compared with placebo, albeit there was an apparent localized reduction in neuronal apoptosis only in the CA3 region of the hippocampus [39]. However, there were reduced numbers of microglia in the gray matter and white matter. Mike et al. [39], found that the caffeine group had higher feeding scores on days 1–3 and improved activity on day 5 compared with the placebo group. However, the groups converged over time, and it is unclear whether the more rapid recovery was related to neuroprotection per se, or to greater alertness mediated by caffeine.

A large study (n = 1,158) in which ewes were given 20 mg/kg caffeine (per os, p.o.) daily in a 4-week lambing period found no improvement in lamb survival or growth rates [41]. This study did not assess neuroprotection. Furthermore, caffeine treatment in a neonatal piglet model of HI was not neuroprotective and was associated with an exacerbation of seizures (Pang and colleagues, personal communication).

There are limited clinical data on the use of caffeine for babies with HIE. In a phase I trial of caffeine for infants with HIE undergoing hypothermia, infants received 20 mg/kg of caffeine followed by two daily doses of either 5 mg/kg (n = 9) or 10 mg/kg (n = 8). This study showed that caffeine administration was safe and did not alter the rate of adverse events [42] compared to the whole-body hypothermia for HIE trial [43]. A retrospective cohort study (n = 52) assessed the safety of the use of methylxanthines, including caffeine or aminophylline, during cooling and showed that it did not affect the frequency of mortality and morbidity, compared with hypothermia only [44]. The efficacy of caffeine was not assessed. Given the lack of data, the safety, efficacy, and optimal dose of delayed, postnatal caffeine for HI brain injury needs to be established in large animal translational studies. If successful, a large randomized controlled trial will be required to determine and efficacy of caffeine for neuroprotection for infants with HIE.

There are well-known safety concerns that high doses of caffeine may have adverse effects. In a pilot study of 74 preterm infants assigned to either standard (20 mg/kg i.v.) loading dose of caffeine citrate or high dose (80 mg/kg i.v.), follow-up at 2 years of age, the high-dose group had a higher incidence of cerebellar injury and alterations in early motor performance [45]. As discussed above, caffeine at high doses can lower the seizure threshold. A secondary analysis of this study showed that the high-dose caffeine was associated with a three-fold increase in seizure duration, compared with low dose [46]. This finding is consistent with the important role of adenosine and the A₁ receptor on seizure control and supports that blocking this mechanism with caffeine could be deleterious.

Additionally, toxic effects of caffeine were shown in a lamb study, with a higher mortality rate in lambs given high-dose caffeine (60 mg/kg i.v. and 30 mg/kg i.v. at 24 and 48 h). There was a rapid but transient drop in blood pressure after caffeine infusion. Although the exact cause of death was unknown, the authors speculated that it may have been related to increased myocardial oxygen consumption, impaired microcirculation, and heart failure. Furthermore, a 120 mg/kg dose of caffeine resulted in 100% mortality in P7 rats after the second dose at 24 h [36].

The timing of treatment is critical to the success of clinical translation and nearly all of the studies discussed in this review administered caffeine either before or immediately after the HI insult (Fig. 2). The few studies of delayed post-HI treatment to date suggest no benefit, as do the large animal studies. If it is to be used as a neuroprotective treatment in infants born with HIE, caffeine administration needs to be tested when started with a clinically realistic delay; if caffeine had a very short window of opportunity that would make it unsuitable for clinical trial. Hypothetically, there may be potential for caffeine to be developed as a prophylactic treatment, given that it can cross the placenta. Specifically, there is growing interest in using caffeine prophylactically in low- and middle-income countries, where infants with HIE are likely born beyond the window of opportunity for treatment with hypothermia. However, we need more information on how long it needs to be given before an HI insult, and how feasible it is to accurately identify at risk pregnancies, especially in low- and middle-income countries where prenatal care is limited. It is unclear why studies in rodents suggest benefit with prophylactic caffeine, before HI, whereas limited benefit with prophylactic caffeine in lambs [39, 47], and deleterious effects of adenosine A1 receptor blockade [20, 47]. Speculatively, this difference may reflect that these rodent studies used inhalational hypoxia and so caffeine may help avoid apnea [47, 48] and so reduce the risk of deep hypoxemia. Regardless of the precise mechanism, these findings strongly suggest that considerable caution is needed before considering human studies.

Next, studies of how caffeine affects the different phases of injury during and after HI are required. In particular, there is potential for adenosine receptor blockade with caffeine to block endogenous neuroprotective pathways that suppress brain activity and limit brain injury as shown by increased injury with selective A₁ adenosine antagonism [20]. As caffeine is a nonselective antagonist of the adenosine receptors, there may be a complex interaction between the positive and negative effects of each receptor subtype. Further, the effect of adenosine receptor blockade on seizure activity needs to be carefully investigated, given that seizures are common after moderate-severe HI. Robust preclinical studies in a variety of large animal models including in fetal sheep/lambs and neonatal piglets (with moderate-severe injury) [49‒51]) that allow for physiological monitoring of EEG, brain swelling, heart rate, and blood pressure [51, 52], as well as neurobehavioral outcomes [53, 54] should be undertaken to understand the effects of caffeine before clinical translation of caffeine for HIE is considered.

It is important to appreciate that large animal models are only established in a few laboratories around the world at present, and longer term recovery is infeasible in most cases. However, such translational studies are essential before considering randomized controlled trials. Next, there needs to be consideration of how to translate the dosage given to animals in preclinical studies to human neonates. As a general rule, most drug dosages cannot be directly compared between animals by weight, but rather should be adjusted by differences in body surface area [55]. For example, taking body surface area into account, a 30 mg/kg dose of caffeine in mouse and a rat is equivalent to considerably lower doses of 2.43 mg/kg and 4.86 mg/kg, respectively, in an adult human [55]. In contrast, maternal administration of caffeine followed by 20 mg/kg loading dose of caffeine to the lamb after birth resulted in lamb plasma levels comparable to the levels targeted in neonates treated for apnea [39]. This highlights the importance of reporting drug plasma concentrations in animal studies and conducting studies in a range of different animals to predict targeted dosing in humans.

Moreover, we were unable to identify any preclinical studies of the combination effect of caffeine with therapeutic hypothermia. It is reasonable to test the potential for neuroprotection when caffeine is given alone, but in high-income countries, infants with moderate-to-severe HIE should receive therapeutic hypothermia and so add-on studies will be essential.

Overall, caffeine seems to have some neuroprotective effects in rodent models of HI brain injury. Even in rodents, the apparent neuroprotective effects were most consistent when caffeine was administered immediately after HI injury. Although the current evidence is limited, it suggests that delayed treatment with caffeine after HI is not effective and may exacerbate seizures secondary to inhibition of the endogenous neuromodulatory effects of adenosine, both during HI and during seizures after HI. Further studies in large animal models to establish a dose response and to monitor physiological parameters continuously are essential, before considering conducting large randomized controlled trials of the efficacy of caffeine for HIE.

The authors have no conflicts of interest to declare.

The authors are supported by the Health Research Council New Zealand (Grant Nos. 17/601, 22/559), the Auckland Medical Research Foundation (Grant No. 3727084), and the Neurological Foundation (3729609 and 3727809).

Conceptualization: K.Q.Z. and J.O.D.; writing – original draft preparation: K.Q.Z.; writing – review and editing: K.Q.Z., F.L., L.B., A.J.G., and J.O.D. All authors have read and agreed to the published version of the manuscript.