Caffeine is often available as caffeine citrate, which comes in both oral and injectable formulations, and the dose of caffeine base is half that of caffeine citrate (Shrestha and Jawa, 2017). As early as in 1977, Aranda et al. published the first study of caffeine used to treat AOP (Aranda et al., 1977). In that study, 18 preterm infants received an intravenous loading dose of 20 mg/kg caffeine citrate followed by a maintenance dose of 5–10 mg/kg once or twice daily for 2–3 days, and a marked reduction in apnea spells was observed. In the next 10 years, the same dose regimen was tested in several studies with small sample sizes (n = 16 to n = 23), and the therapeutic effect of caffeine on AOP was observed by comparison with placebo or theophylline (Murat et al., 1981; Brouard et al., 1985; Anwar et al., 1986; Bairam et al., 1987). In 1999, a multicenter, double-blind, randomized trial of caffeine citrate was performed using the above dose regimen. In this trial, eighty-five infants who were 28–32 weeks post-conception and 24 h or more after birth were randomized to caffeine or placebo for up to 10 days, and the results showed that this dose regimen was safe and effective for those recruited neonates (Erenberg et al., 2000). Based partly on such data, the U.S. Food and Drug Administration approved the dose regimen of caffeine citrate as a loading dose of 20 mg/kg followed by an intravenous or oral maintenance dose of 5 mg/kg/day, which is similar to what was approved by the European Medicines Agency (Erenberg et al., 2000; NDA 20-793/S-001, 2000; European Medicines Agency, 2009). Therefore, in this review, we refer to this as the “standard dose” regimen for caffeine.

In 2006, a large, multicenter, randomized, placebo-controlled trial, called the CAP trial, revealed the short- and long-term efficacy and safety of the standard dose regimen of caffeine (Schmidt et al., 2006; Schmidt et al., 2007). In the CAP trial, preterm infants with very low birth weight (VLBW, 500–1,250 g) were randomized to placebo or caffeine citrate at a loading dose of 20 mg/kg, followed by a maintenance dose of 5 mg/kg/24 h, which could be increased to 10 mg/kg/24 h for persistent apnea. This trial demonstrated several well-known beneficial short-terms effects of caffeine (Schmidt et al., 2006). Regarding the long-term effects, preterm infants had a higher rate of survival without neurodevelopmental disability and a lower incidence of severe ROP, cerebral palsy and cognitive delay at a corrected age of 18–21 months (Schmidt et al., 2007), with an improvement in gross motor function at 5 years (Schmidt et al., 2012). In addition, they also revealed that neonatal caffeine therapy at the doses used in CAP trial is effective and safe into miiddle school age (Doyle et al., 2017; Schmidt et al., 2017; Murner-Lavanchy et al., 2018; Schmidt et al., 2019). Due to the CAP trial, the standard-dose caffeine regimen has been widely used (Table 3). However, variable clinical outcomes do exist after standard-dose caffeine treatment.

Many studies have shown higher doses of caffeine to be more effective with negligible adverse effects (Table 4). Multiple studies have reported that higher doses of caffeine are more effective in reducing episodes of apnea and reducing extubation failure rates (Scanlon et al., 1992; Mohammed et al., 2015; Zhao et al., 2016; Wan et al., 2020). Among them, Mohammad et al. compared a higher dose (loading 40 mg/kg and maintenance of 20 mg/kg/day) with standard-dose caffeine citrate in 120 preterm infants < 32 weeks gestation with AOP within the first 10 days of life (Mohammed et al., 2015). In this trial, the higher dose of caffeine, in addition to being observed to have a better therapeutic effect, was also associated with a significant increase in tachycardia episodes. However, the clinical findings in this trial had no significant impact on physicians’ decision to withhold caffeine.

Other RCTs examined different dosing regimens of caffeine citrate for periextubation management of ventilated preterm infants. In 2003, Steer et al. compared three dose regimens of caffeine citrate (3, 15 and 30 mg/kg) for periextubation management of 127 preterm infants < 32 weeks gestation who were ventilated for > 48 h and found that there was no statistically significant difference in the incidence of extubation failure between different dosing groups (Steer et al., 2003). However, in a subsequent multicenter, double-blind RCT, the same authors found that a dose of 20 mg/kg was given 24 h before a planned extubation or within 6 h of an unplanned extubation reduced the rate of extubation failure within 48 h compared to a lower dose of 5 mg/kg, without evidence of harm in the first year of life (Steer et al., 2004). With the inclusion of additional subjects in the above-mentioned study, Gray compared the long-term effects of the two dose regimens used in Steer’s study (Gray et al., 2011). In this trial, 20 mg/kg/day caffeine citrate resulted in neither adverse outcomes in cognitive development, temperament, morbidity, mortality or disability at 1 year nor in behavior at 2 years.

Several findings also revealed the benefits of higher doses of caffeine for VLBW preterm infants. A retrospective analysis suggested that a higher average daily dose of caffeine citrate was associated with better neurodevelopmental outcomes in VLBW infants (Ravichandran et al., 2019). Another RCT trial found that a maintenance dose as high as 10 mg/kg better reduced the duration of apnea and caffeine treatment in this population (Zhang et al., 2019).

Four meta-analyses also synthesized the findings from the trials comparing higher and lower doses of caffeine citrate (Table 5). Among of them, three papers reported that higher caffeine dosage regimens might be better in reducing the risk of BPD and extubation failure (Chen et al., 2018; Pakvasa et al., 2018; Vliegenthart et al., 2018; Brattström et al., 2019), two reported a decrease in apnea frequency (Chen et al., 2018; Brattström et al., 2019) and one reported a shortened duration of mechanical ventilation (Brattström et al., 2019). Regarding safety concerns, three meta-analyses concluded a higher risk of tachycardia with higher dose of caffeine (Chen et al., 2018; Pakvasa et al., 2018), but no other adverse outcomes were increased.

However, a pilot RCT found an increased incidence of cerebellar hemorrhage (CBH) in infants < 31 weeks’ gestation who were randomized to a higher-dose caffeine citrate (loading 80 mg/kg) (McPherson et al., 2015). Further analysis of this trial demonstrated that early high-dose caffeine therapy was associated with a trend toward an increase in seizure incidence (40 vs 58%, p = 0.1) and burden (48.9 vs 170.9, p = 0.1) (Vesoulis et al., 2016). These results discouraged a larger RCT. More recently, a retrospective study of 218 preterm infants < 28 weeks’ gestation who received a loading dose of caffeine citrate within the first 36 h of life was conducted (Firman et al., 2019). The use of early high loading dose caffeine citrate (a median dose of 80 mg/kg) was not shown to be associated with CBH. Although the two studies obtained different short-term outcomes, they both found that at 2 years of age, the Bayley-III scores used to assess neurodevelopment were not significantly different between the two dose groups.

Collectively, most previous RCTs had small sample sizes, and only two of them have reported 2 years clincal outcomes, which, although positive, need to be treated with caution. Thus, whether to use higher caffeine dosage regimens and how to optimize the caffeine dose are still questionable.

The role of TDM for the control of therapeutic ranges of caffeine has often been challenged due to its benign safety profile when standard dosing is used. As early as in 1977, Aranda et al. revealed that the plasma concentration of standard dose caffeine needed to be monitored and the effective therapeutic concentration was established at 5–20 mg/L by referring to the use of theophylline (Aranda et al., 1977). Subsequently, the same authors also noted that the minimum effective plasma concentration of caffeine was 3–4 mg/L, but an optimal ventilatory response was observed at greater than 8 mg/L, and slight toxicity manifesting as temporary jitteriness was not detected until 50–84 mg/L (Aranda et al., 1979a; Aranda and Turmen, 1979). Therefore, Aranda et al. concluded that the optimal therapeutic concentration of caffeine is 8–20 mg/L, which both produces an adequate response to control apnea and avoids the risk of toxic effects (Aranda and Turmen, 1979).

Blood caffeine levels in preterm infants were almost within this conventional target range in other studies using similar standard dose regimens. In an RCT, 37 preterm infants rapidly achieved the therapeutic concentration within 24 h after starting treatment with a significant reduction in apneic episodes (Romagnoli et al., 1992). A study of 18 Asian preterm infants reported mean serum caffeine concentrations of 10–20 mg/L, and concluded that conventional caffeine therapeutic concentrations should be adhered to in order to ensure safety and efficacy (Lee et al., 2002). Leon et al. found that when the maintenance dose was 6 mg/kg, the 25th to 75th percentile range of mean serum caffeine concentrations in 108 preterm infants was comparable between two different loading dose groups (20 or 25 mg/kg), ranging from 18 to 23 mg/L (Leon et al., 2007). Another study found that the majority of preterm infants achieved target plasma caffeine levels of 5–20 mg/L when treated with a median dose of 5.0 mg/kg (range 2.5–10.9 mg/kg), with 95% of measures within this range in a cohort of 101 preterm infants with 23–32 weeks gestation, including those with renal or hepatic dysfunction (Natarajan et al., 2007a).

Therefore, blood caffeine concentrations of 5–20 or 8–20 mg/L have been commonly recognized as effective therapeutic concentrations for AOP treatment. Routine monitoring of caffeine levels is not recommended by the American Academy of Pediatrics Committee on Fetus and Newborn in their statement on AOP (Eichenwald, 2016). However, when we traced back to the origin, we recognized that the study by Aranda et al. was the fisrt study to determine the therapeutic concentration range of caffeine only based on 18 premature infants’ data (Aranda et al., 1977). Surprisingly, the blood caffeine concentrations were not measured in the well-known CAP trial and the drug was monitored according to its clinical effect only (Schmidt et al., 2006). Of note, the study by Natarajan et al. included a group of preterm neonates (n = 94) who lacked clinical response and had median to 75th quartile of plasma caffeine concentrations of 10.2–14.1 mg/L, suggesting that some neonates may need higher targets of caffeine to control apnea (Natarajan et al., 2007a). Collectively, whether to monitor the level of caffeine in preterm neonates using standard doses still needs to be explored.

Many studies have shown that using higher dose of caffeine was more effective with negligible adverse effects than the standard-dose regimen and explored different effective therapeutic ranges of caffeine. A caffeine PK study including 13 premature infants found that the blood caffeine level varied widely from 12 to 36 mg/L when the single dose regimen of 15 mg/kg was used (Gorodischer and Karplus, 1982). Another RCT reported that 73% of the plasma caffeine concentration measurements in the high-dose group ranged from 26 to 40 mg/L, and apnea episodes were reduced more rapidly within 8 and 24 h without serious adverse effects compared to the standard-dose group (Scanlon et al., 1992). In a PK study conducted by Lee et al., in which no undesired consequences occurred when the mean serum caffeine concentrations were 35.8 or 69.0 mg/L (Lee et al., 1997), a therapeutic concentration > 35 mg/L was proposed to effectively prevent apnea after extubation. Similarly, Steer et al. reported that two higher dose groups with mean serum caffeine concentrations of 31.4 and 59.9 mg/L had short-term benefits and safety during peri-extubation among 127 infants < 32 weeks gestation (Steer et al., 2003). Subsequently, a commentary by Dr. Gal in 2007 questioned the traditional therapeutic concentration (Gal, 2007). According to his findings, higher serum caffeine concentrations produced more significant clinical responses including the reduced incidence of apnea, bradycardia, and of oxygen desaturation, which affirmed a target range of 8–40 mg/L, proposed by Natarajan et al. in another review (Natarajan et al., 2007b). In addition, a retrospective chart review of 198 infants born ≤ 29 weeks gestation showed that serum concentrations of caffeine > 14.5 mg/L were correlated with a reduction in the incidence of chronic lung disease (Chavez Valdez et al., 2011).

However, a small observational prospective study found that serum caffeine levels ≥ 20 mg/L were associated with increased proinflammatory cytokines in preterm infants during the first week of life (Alur et al., 2015). In another study of 115 preterm infants, there was no association between episodes of apnea and serum caffeine concentrations, although there was a significant but weak correlation between caffeine concentration and heart rate (Yu et al., 2016). Meanwhile, some cases reported acute intoxication due to overdose. A case report in 1980 presented two full-term infants with acute caffeine overdose who still had seizure activity when caffeine levels decreased to 31.9 mg/L and 10 mg/L, respectively, although the effect of perinatal asphyxia could not be ruled out (Banner and Czajka, 1980). Another 31 weeks gestational neonate experienced toxic reactions, including hypertonia, sweating, tachycardia, heart failure, pulmonary edema, metabolic disturbances and gastric dilatation, due to the blood caffeine level’s reaching 217.5 mg/L at 36.5 h after dosing, but these symptoms disappeared on day 7 at plasma concentrations of 60–70 mg/L (Anderson et al., 1999). Neurological symptoms, such as uninterrupted tremors, hypertonia, persistent reflex posture, crying, and digestive disorders were reported in a 33 weeks preterm newborn with a serum caffeine level of 160 mg/L at 66 h after administration, whereas his psychomotor development returned to normal after 3 months of age (Perrin et al., 1987). In addition, it is unfortunate that blood caffeine concentrations of subjects were not provided in most RCTs investigating doses of caffeine for AOP (Gray et al., 2011; McPherson et al., 2015; Mohammed et al., 2015; Zhao et al., 2016; Zhang et al., 2019; Wan et al., 2020). Due to lack of high-quality evidence for the long-term safety of high levels of caffeine, further determination of the therapeutic concentration range is difficult.

In the aforementioned studies, thetherapeutic concentration of caffeine was commonly recognized as 5–20 or 8–20 mg/L when using the standard dose regimen. However, some preterm neonates lacked a positive clinical response, although their caffeine levels were within the therapeutic concentration range, suggesting that these neonates may need to use higher doses to control apnea episodes. But using high doses may induce adverse reactions, and how to determine therapeutic doses for neonates who lack a clinical response still needs to be investigated. Therefore, is it feasible to guide dose optimization based on the monitoring caffeine levels?

Refer to Therapeutic Concentration of Standard Dose of Caffeine, routine monitoring of blood caffeine levels is generally not recommended. Leon et al. found that when a caffeine dose regimen close to standard (loading 20 or 25 mg/kg and maintenance of 6 mg/kg/day) was used, the serum drug concentrations were maintained in a safe therapeutic range and were independent of corrected gestational age, weight, and postnatal age within the first 2 weeks of life (Leon et al., 2007). Nevertheless, a PPK study found that the day-to-day variability in caffeine clearance of preterm neonates was twice the interindividual variability, implying that adjusting maintenance doses in light of previous serum concentrations is futile (Charles et al., 2008). However, some studies reported that higher levels of caffeine resulted in a greater response, and caffeine concentration monitoring was essential to ensure reaching the expected drug levels (Gal, 2007; Kahn and Godin, 2016). The 2019 guidelines of the National Institute for Health and Care Excellence recommend that caffeine levels should be monitored using reference ranges from the local laboratories to ensure safety when the daily maintenance dose is higher than 20 mg/kg (NICE, 2019). Combined with clinical practice, a growing body of research has endorsed the view that therapeutic monitoring of caffeine is of interest when therapeutic response is lacking or toxicity is suspected (Natarajan et al., 2007a; Gal, 2007; Leon et al., 2007; Gal, 2009; Kahn and Godin, 2016; Yu et al., 2016).

Naturally, the dose optimization of caffeine cannot be generalized. On the one hand, the change in caffeine clearance in preterm infants is a postnatal maturational progression (Aranda and Beharry, 2020). For routine use of caffeine, Koch et al. developed a simulated PK model and proposed an adjustment strategy based on postnatal age to maintain stable caffeine concentrations, with steps of increasing the caffeine maintenance daily dose by 1 mg/kg every 1 to 2 postnatal weeks, 6 mg/kg in the second week, 7 mg/kg in the third to fourth weeks, and 8 mg/kg in the fifth to eighth weeks (Koch et al., 2017). Recently, it has also been proposed that individualized caffeine medication can be administered with the help of a physiologic based pharmacokinetics (PBPK) model (Abduljalil et al., 2020; Aranda and Beharry, 2020; Verscheijden et al., 2020). On the other hand, the clinical response is specific to each individual and influenced by many factors such as gestational age, birth weight and genetic variability (Gal, 2007; Bloch-Salisbury et al., 2010; Francart et al., 2013; Ravichandran et al., 2019; He et al., 2020). Although increasing evidence has proven that the higher dose of caffeine is beneficial for newborns, there are also potential toxic risks and unknown long-term safety problems. In addition, the reported therapeutic concentration ranges of caffeine may not be simply combined together because of the differences such as the population, sample size, biological matrix, as well as assay methods in each study. This highlights the need to tailor the most appropriate range of individual therapeutic concentration according to blood caffeine levels, and the development of minimally invasive sampling techniques and noninvasive sampling of caffeine may contribute to achieving this requirement (Patel et al., 2013; Bruschettini et al., 2016; Chaabane et al., 2017).