• hypoxaemia
• respiratory illness

Drs Huicho, Singi, and Bharti make the important points that definitions of hypoxaemia should be based on altitude-specific normal values and that further research at sea level and higher altitudes is needed. An altitude-specific definition of hypoxaemia (being an arbitrary value of SpO₂ more than 2¹ or 3 standard deviations below the normal population mean) may be different from the threshold SpO₂ for giving oxygen. Other considerations for giving oxygen are at what level of SpO₂ (at different altitudes) oxygen is beneficial, local resource availability, and, in an individual child, confounding factors including the duration of exposure to altitude, age, or co-existent disease such as brain injury, severe anaemia, pulmonary hypertension, and cardiac failure.

We studied Papua New Guinean neonates and children living at an altitude of 1600m to determine normal range of oxygen saturation.² Hypoxaemia in our study was a SpO₂ more than 2SD below the mean. In practice our threshold for giving oxygen to sick children (SpO₂<85%: more than 3SD below the mean) was lower than this because of limited oxygen availability. However there is evidence that this is safe and effective.³ We stated that without further evaluation this should not be applied to hospitals at substantially lower altitudes than 1600m or in areas where oxygen availability is greater.

In comparing the prevalence of hypoxaemia between studies in different health facilities referral and selection biases are likely. Hypoxaemia will be more common in emergency departments of referral hospitals than at primary care settings, and more common still among children requiring hospital admission.⁴ The prevalence of hypoxaemia in hospitalised children will depend on thresholds for admission and case-mix. The 491 children in our study constituted about 20% of all the children admitted during the course of the study. A specialist paediatrician, whose practice was to oversee the care of sicker children, enrolled many of the patients, so this was a further source of selection bias. The much lower overall prevalence of hypoxaemia seen by Drs Singhi and Bharti in their emergency department population is therefore understandable. Of note the prevalence of hypoxaemia among sick neonates admitted to Goroka Hospital (43%) was similar to the prevalence among young infants (<2 months of age) attending the emergency department in Chandigarh (38.5%).⁵

It is interesting to consider the effects of altitude on hypoxaemia in children with pneumonia. Some populations living at higher altitudes have a greater tendency to pulmonary hypertension; this susceptibility may be genetically determined⁶ and supports Dr Huicho’s statement that ethnic differences in SpO₂ at the same altitude are important. At altitude in response to hypoxaemia, pulmonary blood flow is shunted to the lung apices associated with an exaggerated vasoconstriction in the basal lung.⁷ This may have an adverse effect on ventilation perfusion matching in the supine position. In addition cardiac expression of natriuetic peptides increases in parallel with pulmonary artery pressure.⁸ These and other pathophysiological changes may account for the greater severity and prolonged duration of hypoxaemia seen at higher altitudes.^3,9,10 It may be useful to evaluate the simple intervention of nursing children with pneumonia and hypoxaemia at high altitude in an inclined head-up position, rather than supine, to determine if this reduces the severity of hypoxaemia. There is a need for more evidence about the prevalence of hypoxaemia at sea level and different altitudes; which children benefit from oxygen; for how long oxygen should be given and the best ways to deliver oxygen in remote settings. Controlled trials of oxygen in mild hypoxaemia may not be justified for ethical reasons, but other evidence will be informative. Before the introduction of pulse oximetry in Goroka we used the World Health Organization guidelines for giving oxygen (cyanosis, inability to feed or severe respiratory distress). With the introduction of pulse oximetry we set a threshold for giving oxygen at SpO2 85%. The severe pneumonia case-fatality rate fell from 10% (26 / 258) pre-pulse oximetry to 5.8% (65/1116) 2 years later.^3,10 In highland PNG children cyanosis was only detected in 44% of those with an SpO₂ 70–84%.³ Although there will be confounders in the before-and-after analysis of outcome, we conclude that clinical signs must miss a significant proportion of children who would otherwise benefit from supplemental oxygen, and adherence to a protocol for the administration of oxygen based on a threshold SpO2 of 85% (more than 3 SD below the mean for normal children in Goroka) resulted in improved outcomes, and was within available resources.

The costs of oxygen and logistics of transporting cylinders are major problems in many developing countries; Dr Huicho is right that these are important public health challenges. They call for innovative research and development into how best to supply oxygen to children who need it. The role of oxygen concentrators need to be further explored;¹¹ the combination of concentrators with pulse oximetry would be appropriate technology for many hospitals in developing countries. Increasing the availability of any drug that is crucial to the management of more than 20% of children hospitalised worldwide should be a very high priority; oxygen is one such drug.