We read with interest and welcome the above paper by Mongelli et al.1 (Vol 104, October 1997). In agreement with a previous publication*, Mongelli et al. concluded that there were no significant differences in fetal heart rate (FHR) baseline estimation by experts, defining baseline as a single value. However, it is not possible to reach this conclusion based on the 95% confidence interval (CI) for the differences between experts of −10 to +10 bpm, mentioned in the paper. Indeed, with the presumed number of agreement trials included in the study (n = 3960, for 12 experts and 60 traces), the reported 95% confidence interval would suggest impossible differences between experts as the standard deviation would be 320 bpm. Therefore, it should be clarified if the interval −10 to +10 bpm represents the 95% CI for the paired differences between experts or, as suggested by Fig. 3 in the paper, the Bland and Altman's limits of agreement, which were mentioned in the methods but not in the results.
As clinicians, we congratulate Mongelli et al. for their effort regarding the description of an algorithm for baseline estimation more closely related to a pathophysiological background rather than to a pure mathematical definition, but problems may still arise from their option. Indeed, it has already been shown that the modal FHR value, or any approximated modal value, cannot correctly describe the baseline in largely accelerative and decelerative patterns (eg, with >80% of FHR values above or below the baseline), such as pattern D and some second stage of labour patterns1. If this is not recognized, rare but important clinical errors can occur (eg, by interpreting a normal largely accelerative pattern as an abnormal tachycardic and largely decelerative one4. We regret that alternative pathophysiological approaches to baseline estimation2–3 did not receive more attention from those concerned with computerised analysis of cardiotocograms.