Article Text

Download PDFPDF

High versus standard dose caffeine for apnoea: a systematic review
  1. Roos Vliegenthart,
  2. Martijn Miedema,
  3. Gerard J Hutten,
  4. Anton H van Kaam,
  5. Wes Onland
  1. Department of Neonatology, Emma Children’s Hospital, Academic Medical Center, Amsterdam, Noord-Holland, The Netherlands
  1. Correspondence to Roos Vliegenthart, Department of Neonatology, Emma Children’s Hospital, Academic Medical Center, Amsterdam, The Netherlands; %E2%80%83%E2%80%83%E2%80%83%E2%80%83%E2%80%83%E2%80%83%E2%80%83%E2%80%83r.j.vliegenthart{at}


Background Placebo-controlled trials have shown that caffeine is highly effective in treating apnoea of prematurity and reduces the risk of bronchopulmonary dysplasia (BPD) and neurodevelopmental impairment (NDI).

Objective To identify, appraise and summarise studies investigating the modulating effect of different caffeine dosages.

Methods A systematic review identified all randomised controlled trials (RCTs) comparing a high versus a standard caffeine treatment regimen in infants with a gestational age <32 weeks, by searching the main electronic databases and abstracts of the Pediatric Academic Societies. Studies comparing caffeine to placebo or theophylline only were excluded. Primary outcomes were BPD and mortality at 36 weeks postmenstrual age. Secondary key-outcome was neurodevelopmental outcome at 12 and 24 months corrected age. Meta-analysis was performed using RevMan 5.3.

Results Six RCTs including 620 infants were identified. Meta-analysis showed a significant decrease in BPD, the combined outcome BPD or mortality, and failure to extubate in infants allocated to a higher caffeine dose. No differences were found in mortality alone and NDI. The quality of the outcome measures were deemed low to very low according to the Grading of Recommendations Assessment, Development and Evaluation guidelines.

Conclusions Although this review suggests that administering a higher dose of caffeine might enhance its beneficial effect on death or BPD, firm recommendations on the optimal caffeine dose cannot be given due to the low level of evidence. A large RCT is urgently needed to confirm or refute these findings and determine the optimal dose of caffeine.

  • caffeine
  • apnea of prematurity
  • bronchopulmonary dysplasia
  • dosage regimens
View Full Text

Statistics from

What is already known on this topic?

  • Caffeine is an effective therapy for apnoea of prematurity and

  • It reduces the risk on bronchopulmonary dysplasia and neurodevelopmental impairment. 

  • It is unknown if administering a higher dose of caffeine enhances these beneficial effects.  

What this study adds?

  • Higher caffeine dosage regimens might be beneficial in reducing the risk of death or bronchopulmonary dysplasia.

  • Firm recommendations on the optimal caffeine dose cannot be given due to the low level of evidence.


Apnoea of prematurity (AOP) is a common complication in preterm infants and may be caused by no or insufficient respiratory drive owing to immaturity of the brain stem, obstruction of the (upper) airways, or a combination of both factors.1 2 AOP can result in intermittent hypoxaemia which has been associated with retinopathy of prematurity (ROP) and neurodevelopmental impairment (NDI).3 4 Although invasive mechanical ventilation is the most effective treatment for AOP, it is associated with an increased risk of bronchopulmonary dysplasia (BPD) and NDI.5 Therefore, preterm infants with AOP are first subjected to a combination of pharmacological treatment and non-invasive respiratory support.

Caffeine is a respiratory stimulant belonging to the group of methylxanthines, and has become the standard pharmacological treatment for AOP. Treatment usually consists of a loading dose followed by a maintenance dose (MD). Serum levels of 3–84 mg/L are considered therapeutic and safe.6 The mechanisms of action of caffeine are not completely understood. Reported physiological effects in preterm infants are stimulation of the central nervous system, increased peripheral chemoreceptor responsiveness to CO2,7 a rapid and sustained increase in diaphragmatic activity and an increase in tidal volume.8 Placebo-controlled trials have shown that caffeine reduces the frequency of AOP, increases the chance of successful extubation, reduces the need for MV and reduces the risk of BPD in the preterm infants.7 9 10 In the longer term, caffeine is also associated with an improved neurodevelopmental outcome at 18–21 months corrected age (CA).11 12

Several studies on pharmacodynamics and pharmacokinetics have investigated if a higher dose of caffeine would augment the previously reported beneficial effects without increasing the risk of adverse effects.13 14 The objective of this study was to identify, appraise and summarise studies investigating the effect of a higher versus a standard dose of caffeine in preterm infants.


Criteria for considering studies for this review

Randomised controlled trials (RCTs) comparing a higher versus a standard dosage regimen of caffeine in preterm infants born before 32 weeks of gestation were considered eligible for this review. RCTs comparing caffeine to placebo or caffeine to theophylline only were excluded. To be included in this review, each trial also needed to report at least one of the outcome measures displayed in online supplementary appendix.

Supplementary file 1

Search methods for the identification of studies

Following the guidelines of the Cochrane Neonatal Review Group, two authors (RV and MM) independently identified studies using the Medical Subject Heading terms: ‘Caffeine’ and ‘Newborn’ in the following electronic databases: MEDLINE via PubMed (1996 to March 2017), EMBASE (1980 to March 2017); CINAHL (1982 to March 2017); the Cochrane Central Register of Controlled Trials (CENTRAL 2015, Issue 3) in The Cochrane Library. Also, reference lists of published trials, reviews and abstracts of the Pediatric Academic Societies from 1990 onwards were hand searched. The search strategy had no language restriction. Only RCTs and trials performed in humans were eligible. We searched clinical trials registries for ongoing or recently completed trials.

Data collection and analysis

The relevant citations were divided into three groups by two authors (RV and MM): ‘clearly eligible’, ‘clearly not eligible’ and ‘possibly eligible’. A full-text review was done on all except those classified as ‘clearly not eligible’. Any disagreement was resolved by consensus. Two authors (RV and MM) independently extracted, if applicable, all prespecified outcome measures and the following additional study data: year of publication, region of origin, number of included patients, patient characteristics (birth  weight, gestational age, gender, postnatal age at start of therapy, respiratory distress syndrome, treatment with antenatal corticosteroids and postnatal surfactant), caffeine loading dose, MD of the higher and the standard dosage regimen, duration of therapy and route of administration. The original authors of the included trials were contacted to check whether data extraction was accurate and, if necessary, provide additional data.

The risk of bias of the individual studies was independently evaluated by two authors (RV and MM) according to the Cochrane Handbook for Systematic Reviews of Interventions.15

In addition, the quality of evidence was also assessed for the main outcome measures using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach16 by two independent reviewers (RV and MM), as previously described.17 The following (clinical relevant) outcomes were assessed: combined outcome death or BPD, BPD, tachycardia, necrotizing enterocolitis (NEC), intraventricular haemorrhage (IVH) ≥grade 2, death or disability at 1-year CA. The GRADE results were entered into the Guideline Development Tool, creating a summary of findings table.18 The quality of evidence was assessed as following:

  • High quality: we are very confident the true effect lies close to that of the effect estimate.

  • Moderate quality: we are moderately confident the true effect is likely to be close to the effect estimate, but might also be substantially different.

  • Low quality: we have limited confidence in the effect estimate, the true effect might be substantially different.

  • Very low quality: we have very little confidence in the effect estimate, it might be substantially different from the effect estimate.

Data on similar study outcomes were pooled using Review Manager.19 During data extraction, it became clear that the regimens used in the included studies differed considerably, especially in duration and dose. For example, what one study considered a high-dose therapy regimen, was considered as standard dose therapy in another study.20 21 Given this heterogeneity, we decided to perform two post hoc analyses, categorising the studies according to the duration of difference in caffeine therapy in both arms, that is, shorter and longer than 14 days, and according to the MD in the standard allocation arm, using an arbitrary cut-off of <10 mg/kg/day caffeine-citrate. These cut-off points resulted in a more or less even distribution of studies in both subgroups.

Treatment effect estimates were expressed as typical relative risk (TRR) for dichotomous outcome measures and weighted mean difference for continuous data with a 95% CI. Number needed to treat for an additional beneficial outcome (NNTB) and number needed to treat for an additional harmful outcome (NNTH) were calculated. Degree of heterogeneity between trials was assessed by inspecting the forest plots and using the I2 statistic to quantify the impact of heterogeneity as defined by the Cochrane Neonatal Review Group.22


Results of the search

The search identified 1286 citations (figure 1). Of these, 1221 were clearly not relevant and an additional 58 were excluded, mainly because there was no dose comparison or the design was not a clinical trial. The Gray et al study23 reported the results of all patients of another RCT.24 Only those outcome measures from this study,24 not reported by Gray et al 23 were used for this review. The remaining six RCTs, randomising a total of 620 preterm infants, were included in this review. Detailed description of the studies can be found in the online supplementary appendix. The patient characteristics were similar across trials and the mean postnatal age at which caffeine was started was less than 7 days (table 1). The study of Steer et al 25 had three (high, medium, standard) instead of two (high, standard) study arms, and for this reason we pooled the high and medium group to one (high-dose) group. In all trials, both the loading dosage, as well as the MD of the caffeine citrate varied considerably. Four studies allocated the infants in the standard dosage regimen to an MD <10 mg/kg/day20 21 23 25 and two studies used an MD ≥10 mg/kg/day in the lower dosage group.26 27 Although the total duration of the caffeine in all studies was >14 days, the duration of the therapy difference between the two allocation arms was ≤14 days in two studies.25 26 The trial of Steer et al 25 administered a different caffeine dosage during 7 days, the McPherson trial had a difference in caffeine dosage in the loading dose only, during the first 36 hours of caffeine therapy. The MD in this study was equal in both allocation arms.26

Figure 1

Study selection. PK-PD, pharmacodynamics and  pharmacokinetics; RCT, randomised   controlled trial.

Table 1

Study characteristics

Quality of evidence

The risk of bias in the included studies varied from low to high risk (figure 2). The risk of attrition bias was judged to be high in three studies.20 21 23 The study of Romagnoli et al 20 lacked information about randomisation, allocation concealment and blinding. The study had a high risk of incomplete data, due to a difference in included infants between the text and a table.20 Scanlon et al had a high risk of performance bias, since the intervention was not blinded.21 In this study the loading dose in the high dosage group was given in two separate doses, whereas the loading dose in the low dosage group was given as one dose.21

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study. +: low risk of bias. ?: unclear risk of bias. -: high risk of bias.

The quality of the outcome measures pooled by meta-analysis was deemed low to very low, according to the GRADE guidelines16 due to imprecision and inconsistency of the effect estimates (table 2). Imprecision was caused by wide 95% CI in the meta-analyses due to small sample sizes. Heterogeneity between definitions of high and low dosage of caffeine in different studies caused inconsistency. Also, including studies with high risk of bias lowered the quality of evidence.20 21

Table 2

Summary of findings summarises the quality of evidence and the magnitude of relative and absolute effects for each important outcome

Effects of intervention

An overview of the meta-analyses and forest plots of the pooled outcome measures can be found in table 3 and online supplementary appendix (figure 3A–K), respectively.

Table 3

Meta-analysis of pooled outcome measures

Mortality and BPD

Mortality was only reported at discharge and 12 months CA, and no differences were found between the caffeine groups.

Compared with the infants allocated to the higher caffeine dosage regimen, the infants who were allocated to the standard dosage regimen had no higher incidence of the outcome BPD (figure 3A). The subgroup analyses of MD (<10 vs ≥10 mg/kg/day) did not alter the effect. However, the subgroup analysis of therapy duration showed a significant effect in favour of infants allocated in the higher dose regimen, when therapy was given for >14 days (TRR 0.72, 95% CI 0.54 to  0.97, NNTB 9, 95% CI 4.7 to 71.0). The combined outcome mortality or BPD at 36 weeks postmenstrual age was reported by three studies and was only significantly different in the subgroup analysis for therapy duration >14 days (TRR 0.76, 95% CI 0.59 to  0.98, NNTB 9, 95% CI 4.7 to  84.6) (figure 3B).

Figure 3

(A) Forest plot of the meta-analysis of bronchopulmonary dysplasia (BPD) at 36 weeks postmenstrual age (PMA). (B) Forest plot of the meta-analysis of BPD and mortality at 36 weeks PMA.

Short-term respiratory outcomes

Five studies reported on the effect of caffeine on apnoea frequency, but differences in the definition of this outcome parameter prevented meta-analysis. Except for the Romagnoli et al study,20 all studies21 23 25 27 reported a significantly lower apnoea frequency in the high-dose caffeine group compared with the standard-dose group.

Failure to extubate was reported less in the infants allocated to the higher caffeine dose (TRR 0.51, 95% CI 0.37 to  0.70; NNTB 7, 95% CI 4.2 to  12.6).

Meta-analysis on duration of respiratory support could not be performed since data were reported in interquartile ranges.23 24 26 27 The individual studies showed no difference in duration of invasive and non-invasive ventilation.23 24 26 27 One study reported significant shorter duration of oxygen therapy in the high dose compared with the standard-dose group (14.5 days vs 20 days, P=0.04).27

Adverse effects

Meta-analysis showed an increased risk of tachycardia for the infants treated with the higher caffeine dose (TRR 3.39; 95% CI 1.50 to  7.64, NNTH 9.1, 95% CI 6.3 to 15.3). There were no differences in the outcomes NEC, spontaneous intestinal perforation, hyperglycaemia, ROP and IVH between the groups. McPherson et al 26 reported a higher risk of focal cerebellar haemorrhage diagnosed with MRI in the high-dose group (36%) versus the standard-dose group (10%) (OR 5.0 (95%CI 1.2 to 20.7)). There were no differences in the incidence of extensive cerebellar haemorrhage.

Long-term neurodevelopmental outcome

The revised Griffiths developmental scale28 was used by one study to investigate psychomotor development at 12-month CA.23 They did not find a difference in these outcome measures between the two groups, except for the outcome general quotient only, favouring high-dose caffeine treatment. The only article reporting data using the Bayley Scale Infant Development III29 at 24 months found no difference between the high and standard dose groups.26


Caffeine therapy in preterm infants reduces the risk on the combined outcome death or BPD and NDI at 5 years CA.7 12 30–32 However, it is unclear if an increased dose of caffeine can safely augment these beneficial effects.33 34 This systematic review shows that the number of trials comparing (head to head) a higher versus a standard caffeine dose are limited. The findings of this review suggest that a higher caffeine dose improves short-term pulmonary outcomes and reduces the risk of the combined outcome measure death or BPD, and BPD alone when administrating this higher dosage for more than 14 days. These findings should be interpreted with caution as only three studies reported this outcome. However, since BPD is nowadays considered a disease of lung development arrest that develops over a longer period of time, this beneficial effect on BPD could be explained by more effective treatment of AOP and a higher rate of successful extubation when administrating a higher dose of caffeine.9 35 36

In contrast to previous publications,36 37 adverse neurodevelopmental outcomes at 12 and 24 months CA were not significantly decreased in the high-dose caffeine group. However, only two studies reported on this outcome measure, with small sample sizes. Due to the inadequate power, small differences might not have been detected. This review also suggests that the aforementioned beneficial effects of administrating a higher dose of caffeine does not come at the expense of more adverse effects. The only consistent adverse effect reported by the different RCTs was an increased risk of tachycardia. Only one study reported an increased risk of cerebellar haemorrhages in the high-dose caffeine group. The clinical implication of this finding on neurodevelopmental outcome remains unclear,38 39 but nevertheless, future studies should closely monitor this possible adverse effect.26

Although the results of this review showed a beneficial effect on the outcomes death or BPD, and BPD alone, the applicability of this review was deemed low for several reasons. First, the overall quality of the outcome measures assessed by the GRADE guidelines16 was deemed low to very low, due to imprecision and inconstancy of the effect estimates. Second, the sample size of all studies was small. Therefore, the studies had inadequate power to detect small but clinical relevant differences. Third, the administered caffeine doses differed considerably between trials and may have hampered the internal validity of our findings. Finally, not all prespecified outcome measures were reported in the studies. Our attempts to retrieve unpublished data were unsuccessful and therefore did not overcome this shortcoming.

A well-designed adequately powered large RCT is needed given the low quality of available evidence and the current variability of clinically prescribed caffeine dosage regimens.31 Given the beneficial results of this review administrating a higher MD, the future trial should compare a high MD caffeine citrate (20 mg/kg/day) to a low MD caffeine citrate (10 mg/kg/day) for >14 days in preterm infants at risk for adverse pulmonary and neurodevelopmental outcomes. Although no adverse neurodevelopmental outcomes in the Gray and McPherson trial were found, the trialists of this future study should carefully consider how high the loading dose in the intervention arm should be, given the increased risk of cerebellar haemorrhage in the pilot study of McPherson et al.26


The present systematic review identified, appraised and summarised all available evidence on different caffeine regimens in preterm infants. It shows a potential benefit of a higher caffeine dosage regimen on the incidence of mortality and pulmonary outcomes. However, the low level of evidence and/or paucity of available data prevents firm recommendation on the optimal caffeine dosage regimen for clinical practice. A large multicentre RCT is needed to provide more conclusive evidence on whether higher caffeine dosage regimens improve long-term beneficial pulmonary and neurodevelopmental outcome in preterm infants.


View Abstract


  • Contributors Conception and study design: RV, MM, GJH, AHvK and WO. Collection, analysis and interpretation of data: RV, MM, GJH, AHvK and WO. Drafting the manuscript for important intellectual content: RV, MM, GJH, AHvK and WO. Decision to submit the paper for publication: RV, MM, GJH, AHvK and WO. RV wrote the first draft of the manuscript. All authors approved the final version to be published. RV will act as corresponding author for this paper.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Fantoms
    Martin Ward Platt