We read with interest the paper Analysis of intrapartum cardiotocography by Chung et al. (Vol 102, June 1995). The authors present a simple algorithm (325 lines of code) for computerised CTG analysis and conclude that it is capable of predicting acidosis at delivery. This claim cannot pass without comment.

Their definition of acidosis (pH cord artery, pHa < 7.15) is no longer accepted as clinically significant. Our work2 and that of others3–5 clearly shows that a pHa < 7.05 is a more appropriate definition of acidosis which corresponds to the 2.5th centile of the population, whereas a pHa i 7.15 corresponds to the 10th centile. In their study there was only one case with cord artery pH < 7.05. If the results are interpreted using this lower cut-of then 22 of the 23 cases that were identified as “abnormal” were not acidotic. Similarly the authors have used an unacceptable value for Base Excess (BE < − 8) which should more appropriately be BE < −12 to indicate significant metabolic acidosis. Any study which relies on an assessment of perinatal outcome must include both cord artery and vein pH and Base Deficit (or excess) of the extracellular fluid (not blood) and must also include Apgar scores, as well as any relevant clinical factors such as need for resuscitation neonatal hypoglycaemia and abnormal neonatal behaviour. The classification criteria used to identify an abnormal CTG are poorly refined. Are the authors really suggesting that any tachycardia > 160 or bradycardia < 110 or the presence of five late decelerations in the entire course of labour constitutes an abnormal CTG predictive of acidosis?

There are also some serious problems with the signal processing employed. The system filters the beat to beat heart rate and then samples the filtered signal at 10 second intervals, providing six points per minute. This smooths the signal as Figure 1 shows and makes it impossible to assess heart rate variability, which is key to any assessment of the CTG. They justify this approach in terms of a need to reduce the amount of data. However, if the fetal heart rate were as high as 200 bpm then there would be just 3.33 samples per second of raw data to process. This is trivial for modern computers the compact disc player processes over 44000 samples per second.

We do not believe (and the authors have not shown) that their simple algorithm can predict a significant level of acidosis and we find it extraordinary that this paper was published with such a title. We believe that the CTG requires sophisticated interpretation in the context of clinical information. We have recently shown that when this complete clinical picture is provided, experts can agree and be consistent in their management of labour. Furthermore, a computer system which also considers this complete information can achieve a performance indistinguishable from experts6.


  1. Top of page
  2. References