Response to Duke et al

• hypoxaemia
• respiratory illness

We read with interest the article by Duke et al¹ regarding hypoxaemia in acute respiratory and non-respiratory illnesses in infants and children in developing countries published recently in Archives.¹ The authors have rightly pointed out the limited availability of published data on the incidence, significance or clinical signs predicting hypoxaemia in infants less than three months of age. With similar concerns we had conducted a study in infants less than two months, a part of which was published in the Archives.² We found that tachypnoea, defined as RR>60/min, predicted hypoxia with 80% sensitivity and 68% specificity.² In that study we had also examined six functional and behavioural responses as predictors of hypoxemia (table 1). Five of these six variables had a very good sensitivity to detect hypoxaemia.

A very high prevalence of hypoxaemia in the population studied by Duke et al is rather intriguing. Out of total 257 sick neonates and children 52%, were hypoxaemic. Among children with acute lower respiratory infection (ALRI) 73% and those with non-ALRI 32% were hypoxaemic. In an ongoing study we have measured oxygen saturation (by Nellcore™ oximeter) in a prospective cohort of 683 children 2–59 months brought to paediatric emergency department (ED) with any respiratory symptom. Oxygen saturation using a fingertip sensor in these children at the time of arrival to ED ranged from 78–99%. The overall prevalence of hypoxaemia defined as SpO₂ <90% was 4.5% (table 2).

An additional 5.1% children had borderline hypoxaemia, i.e. a SpO₂ value of 90%. This is similar to a prevalence of 5.9% hypoxaemia (defined as SpO₂ <90%) in Gambian children, 2–33 months of age, reported by Usen et al.³ Even in our previous study of 200 infants less than two months, only 38.5% of the sick infants attending ED were hypoxaemic.² A systematic review of studies on prevalence and predictors of hypoxemia in children by Lozano et al⁴ found that the prevalence of hypoxia was dependent upon a number of factors including the setting of the study. The prevalence ranged from 6–9% in outdoor setting to 31–43% in emergency departments to a maximum of 47% in hospitalised children.

Yet, in our study, which represents the situation near sea level (Chandigarh being a plain topographically) and the setting of an emergency department, the prevalence of hypoxaemia is much lower than that reported at heights. In light of our data and published literature. We believe that either the definition of hypoxemia used by Duke et al¹ is too liberal or the children with respiratory symptoms living at high altitude decompensate more frequently to develop hypoxia. More information is needed in this respect to formulate guidelines for general use. The cumulative data clearly suggest that hypoxaemia is more frequent in children living at high altitude. Interestingly most studies including that of Duke et al on this subject in children 2 to 59 months have been from high altitudes. It is most likely that geographic location, 1600m above sea level is responsible for the high frequency of “hypoxaemia” in their patient population. This, however, may not necessarily reflect the need for oxygen therapy. If definition of hypoxemia suggested by Duke et al were to be applied as a guideline to oxygen therapy almost half of their patients would need oxygen therapy. We need to answer as to whether oxygen therapy makes any difference to outcome of patients labeled as hypoxaemic using cut off limits proposed by Duke et al. It may also be worthwhile to conduct studies with a large sample size at sea level (plains) and in various settings before reaching a conclusion about SpO₂ cut off for hypoxia at heights.