The use of methylxanthines to reduce the frequency of apnoeic episodes in premature infants was first reported more than 35 years ago. Since then, and despite a lack of data addressing long-standing concerns about potential risks, methylxanthines have become such a routine part of the clinical treatment of apnoea of prematurity that caffeine, the methylxanthine of choice due to its long half-life, ranks as one of the 10 drugs most commonly prescribed in neonatal intensive care. The most cited concerns associated with therapeutic levels of caffeine include that reductions in cerebral blood flow and inhibition of adenosine receptors, which protect against energy failure and cell death during experimental hypoxia, might compromise the hypoxia tolerance of brain cells and adversely affect nervous system development. A caffeine-induced elevation of metabolic rate (∼20%) at a time when nutritional intake is already compromised may also adversely affect growth (Schmidt et al. 2007; Stevenson, 2007).

Despite these concerns, data are emerging from the Caffeine for Apnea of Prematurity Trial that at an age of 18–21 months (corrected for premature birth) the benefits of caffeine therapy in infants with very low birth weight outweigh the risks. Caffeine therapy improved the rate of survival without neurodevelopmental disability; death rates, deafness, blindness and growth were unaffected; incidence of cerebral palsy and cognitive delay were reduced (Schmidt et al. 2007).

While these data are encouraging, animal studies suggest that the negative impact of caffeine also results from altered development of the adenosinergic signalling system and subsequent disruption of adenosine-modulated behaviours (Pan & Chen, 2007). The potential long-term impact of neonatal caffeine treatment on adenosine-modulated behaviours is elegantly demonstrated by Montandon et al. (2009) in this issue of The Journal of Physiology. Using telemetry and whole-body plethysmography to minimize animal disturbance and ensure accuracy of behavioural measures, the authors demonstrate in rats that a regimen of orally administered caffeine (15 mg kg⁻¹ day⁻¹) from postnatal day 3 to 12 changes sleep architecture, and alters respiratory control system activity in adulthood (8–10 weeks). Given the importance of adenosine in sleep, this analysis was long overdue; effects were dramatic. Caffeine-treated animals had a 50% greater latency to sleep onset, spent more time in wakefulness (51 vs. 32% in control), less time in non-REM sleep (44 vs. 61% in control), and despite the overall reduction in non-REM sleep, had almost 50% more non-REM sleep episodes (i.e. sleep was more fragmented). Resting ventilation was also elevated and ventilatory responses to increased CO₂ were reduced in all states.

The direct clinical relevance of rodent data is always uncertain. However, they do reveal the potential for long-term disruption of sleep pattern and homeostatic control systems (Montandon et al. 2009). Thus, in light of the benefits of caffeine therapy for apnoea of prematurity at 18–21 months (Schmidt et al. 2007), the question of how to balance treatment gains against possible risks will remain until comparable human data are available. Data for the 5 year follow up of the Caffeine for Apnea of Prematurity Trial are pending, but are unlikely to include assessments of sleep, respiratory control or other adenosine-modulated behaviours. It will be some time before such adult data will become available from this cohort.

What can be done in the meantime to maximize gain and minimize risk of caffeine therapy? Effective strategies require a mechanistic understanding of caffeine's beneficial and negative actions. Thus, a priority will be to gain this level of understanding, to which end animal studies will be invaluable. Physiological actions of methylxanthines include increased minute ventilation, improved CO₂ sensitivity, decreased hypoxic depression of breathing, enhanced diaphragmatic activity, and decreased periodic breathing. Indeed, Schmidt et al. (2007) attributed ∼50% of the beneficial actions of caffeine therapy to reduced time on positive airway pressure ventilatory support or supplemental oxygen. It is assumed, although not established, that a reduction in apnoeic episodes equates to a reduction in hypoxaemic episodes and that some benefit is derived from avoiding negative consequences of chronic intermittent hypoxaemia. Despite this list of effects, the pharmacological basis for the actions of caffeine is not clear. In its therapeutic range (8–20 mg l⁻¹) competitive antagonism of A1 and A2a receptors is likely (Comer et al. 2001), while inhibition of phosphodiesterase activity is unlikely. A1 receptors inhibit central respiratory networks, at least in the early perinatal period, while activation of A2a receptors can inhibit respiratory activity by activating medullary GABAergic neurons (Abu-Shaweesh & Martin, 2008). Block of these pathways is likely to contribute to the stimulation of breathing but other actions are possible. It has also been suggested that caffeine, through A1 receptor antagonism, protects against hypoxia-induced white matter injury (Abu-Shaweesh & Martin, 2008).

The potential negative effects of caffeine therapy are also poorly understood mechanistically. The study by Montandon et al. (2009) should stimulate research in this area. Important objectives will include: determining critical periods of sensitivity and how these relate to human development; defining the pharmacological basis of positive and negative effects, since resolving underlying receptor mechanisms will provide cues to help decipher (and eventually manipulate) specific signalling cascades; as well as developmental studies to determine the stage at which caffeine-mediated changes in behaviour emerge – those that manifest early will be more amenable to examination in parallel human studies. In the end, while the intriguing finding of Montandon et al. in rodents means that researchers and clinicians are likely to lose sleep over balancing the benefits and risks of caffeine therapy for apnoea of prematurity, significant work remains to determine whether infants receiving caffeine therapy will as well.