Authors' Reply



We thank Dr Keith and Professor Greene for their comments on our paper and make the following responses. We acknowledge that a pH of less than 7.05 in the umbilical artery was found to be associated with long-term morbidity. Our intention however was to detect acidosis but not long term morbidity, as the CTG is likely to reflect the physiological status of the baby at the time rather than any cellular damage to the brain and a pH of 7.15 is acidosis. Dr Keith and Professor Greene appear to have misread part of our paper. We did require a continuous tachycardia or bradycardia exceeding 30 minutes. The recursive nature of data processing is such that the time required often increases exponentially to data volume, and many existing system use data averaging as a means of data reduction. The filtered heart rate signal was obtained by passing all detected fetal heart rate through a recursive filter, the output of which was then sampled at ten second intervals. Since beat to beat variability will be lost we measured the variation of the fetal heart rate signal. Periods of low variation corresponded to segments of the trace where accelerations and decelerations were absent. The visual assessment of “variability” is not synonymous with variation which we had defined in our algorithm.

We are presently developing a system sampling the CTG every 2 seconds, and have developed a more complex algorithm, which has in excess of 10000 lines of codes using Object Orientated Programming (C++). This new system overcomes the hardware limitations of the previous study which allowed a maximum data storage of all FECG heart rate variables every ten seconds. This new program measures both periods of low variation as well as the variability in terms of low, medium and high frequency components of the FHR. We are currently evaluating the results of this effort. We would like to point out that we have made the effort to produce a more sophisticated program because the results from this first study showed promise.

We feel that it is a good principle to reduce unnecessary algorithm complexity unless improved performance can be demonstrated. Our paper described a basic algorithm with minimal processing. Its performance can be used as a basis for comparison against more complex and sophisticated algorithms.

We have read with interest the work of Professor Greene and his co-workers. We wish to explore an alternative approach. We prefer to identify the basic components of the diagnostic process and determine their individual relevance to outcome, before testing whether combining them in any particular manner would improve diagnostic precision. Our computerised analysis of the CTG is a step in this approach.

Clinical information has an impact on CTG interpretation and subsequent clinical decision making. The nature and extent of this impact is not known although it is generally assumed to be beneficial. We believe that, in research, evaluating CTG alone and with clinical information are both necessary to define how much clinical information is necessary or helps in the interpretation of the CTG. As the technology for fetal assessment and clinical management may evolve independently, it may also be useful to separate them.