Inter-observer agreement in clinical decision-making for abnormal cardiotocogram during labour: a comparison between CTG and CTG plus STAN
Article first published online: 10 DEC 2009
© 2009 The Authors Journal compilation © RCOG 2009 BJOG An International Journal of Obstetrics and Gynaecology
BJOG: An International Journal of Obstetrics & Gynaecology
Volume 117, Issue 1, page 121, January 2010
How to Cite
Grant, J. (2010), Inter-observer agreement in clinical decision-making for abnormal cardiotocogram during labour: a comparison between CTG and CTG plus STAN. BJOG: An International Journal of Obstetrics & Gynaecology, 117: 121. doi: 10.1111/j.1471-0528.2009.02395.x
- Issue published online: 10 DEC 2009
- Article first published online: 10 DEC 2009
- Accepted 22 July 2009.
There are two main problems with this study.1
1. The authors use as the denominators for the estimation of the 95% confidence intervals the number of trials of agreement. But these trials of agreement are not independent of one another. It is the number of the subjects that should be the denominator. The correct 95% confidence interval for the overall agreement of cardiotocography (CTG) is 0.78 (0.64, 0.93); and for CTG plus ST waveform analysis (STAN) is 0.83 (0.70, 0.97). With the number of subjects as the denominator, none of the comparisons carried out by the authors reach the conventional level of statistical significance. We are not justified in concluding that CTG plus STAN is inferior to cardiotocography alone in the decision to deliver, or is superior to cardiotocography alone in the decision not to deliver.
2. The authors estimate the proportions of agreement for justified intervention and for justified non-intervention, but these estimates are confounded by the design of their study. Ten cases were selected where the infants had a pH at birth of <7.05, and there were 20 cases where the pH at birth was >7.05. In these 20 cases there were eight cases with a pH of 7.05–7.10, still a significant degree of acidosis, and usually indicating delivery. Obviously these were cases indicated by STAN as not indicating delivery. Was the cut-off value between normal and abnormal chosen arbitrarily? Was it chosen post hoc? Why choose pH as the outcome at all, when we know that metabolic acidosis alone is a poor predictor of cerebral palsy?
The preferred method of performing this study would be to select a group of infants with hypoxic-ischaemic encephalopathy, and a group of infants who do not have hypoxic-ischaemic encephalopathy. This is preferable to a random sample of all infants delivered. In well-grown infants at term, progressive fetal hypoxia and metabolic acidosis can occur, until a critical boundary is crossed when cardiovascular collapse results, with consequent spastic quadriplegia. If that critical boundary is not crossed, complete recovery of the infant is the rule, despite severe metabolic acidosis. It is an ‘all-or-nothing’ phenomenon. Therefore, this study should be repeated in a group of infants who have seizures with severe metabolic acidosis, and a group of infants who do not have seizures with severe metabolic acidosis.