Title: Deceleration Area and Deceleration Capacity: Promising
predictors of fetal acidaemia in human labour? Visual versus
computerised cardiotocography
Re: Georgieva A, Lear CA, Westgate
JA, Kasai M, Miyagi E, Ikeda T, Gunn AJ, Bennet L. Deceleration area and
capacity during labour-like umbilical cord occlusions identify evolving
hypotension: a controlled study in fetal sheep. BJOG 2021;
https://doi.org/10.1111/1471-0528.16638.
Dear Editor,
It is intuitive to birth-attendants that the bigger and more frequent
fetal heart rate (FHR) decelerations over a longer period (i.e. bigger
cumulative deceleration area - DA) means increasing chance of fetal
acidaemia/hypotension/hypoxaemic injury. The excellent animal study by
Geogieva et al 1 confirms this expected correlation.
However, obstetricians need to place this in the context of information
already available from well-designed studies in human labour examining
DA.2,3 These studies show that the degree of
correlation (even statistically significant) of DA to neonatal outcomes
does not translate into clinically useful positive/negative predictive
values (PPV/NPV). There are multiple unresolved logistical difficulties
in addition.2,3 The quoted1 study by
Cahill et al mistakenly states that their ‘DA cut-off’ requires five
caesareans to prevent one case of acidaemia; because with its PPV of
4%, the correct calculation is 25 to one. Another analogous
computer-derived parameter the “deceleration capacity (DC)” measures
no capacity, hence somewhat a misnomer. One large cohort study of DC on
22,000 women lacked statistically-significant improved acidaemia
detection and crucially the important receiver-operating-characteristic
(ROC) curves were missing.1 Another cohort study on
11,980 women showed the area under curve (AUC) for DC to be 0.66.4 This
AUC reveals that if we want to detect 90%, 80% or 50% of acidaemic
babies, the same deceleration capacity (DC) was shared by 80%, 60%,
25% of normal babies respectively, making DC disappointing for clinical
application.4 Similarly, The AUCs for deceleration
area (DA) in human studies are very comparable and
wanting.2,3
A lesson from wider experience in artificial intelligence (AI) implies
that a constricted-single-parameter approach (e.g. DC/DA) may be
insufficient for the complex task of intrapartum fetal monitoring.
Computers have been beating chess-grandmasters for 25 years; because
chess offers a “kind learning environment” with fixed rules, patterns
repeating exactingly, feedback extremely accurate and very rapid. In the
“unkind learning environments” devoid of rigid rules, singular domains
and reams of perfect historical data; the AI and machine learning have
been disastrous. Cardiotocography (CTG) requires integration of multiple
FHR parameters with mother-fetus-labour-condition permutations.
Intervention changes outcomes; hence the feedback can be
inaccurate/unreliable. Human cognition assimilates these paradoxes. The
greatest human strength is the exact opposite of narrow specialisation
of AI. It is the ability to integrate broadly.
Research in computerised non-visual parameters like DA or DC like the
current study1 is important but alongside the visual
CTG interpretation which seems indispensable in the foreseeable future.
Computerised interpretation should emulate the visual
pattern-recognition and then supplement it with new parameters when
proven. A concept/philosophy is presented that all FHR decelerations are
due to hypoxaemia.1 It has been suggested that
chemoreflex is an indefatigable guardian of hypoxaemic fetus and
hypoxaemia per se does not matter.5 An argument is
proposed that the classification of decelerations into
early/variable/late is irrelevant (red herring), hence the deceleration
area (overriding the former classification) is the
future.5 Clinical experience so far seems
different.2,3 DA and DC consider all FHR decelerations
to be the same (and hypoxaemic?). Given the complexities/setbacks in
research on intrapartum monitoring, it seems an ambitious challenge to
compute clinically useful thresholds of DA/DC in several different
mother-fetus-labour scenarios. Studies on retrospective
“Big-data” particularly suffer from confounding
factors, importantly cord-gases available on a skewed smaller subgroup.
It seems premature to relinquish to computerisation/DA/DC. Hence,
obstetricians need to preserve and improve the scientific visual CTG
pattern-recognition given the limitations, changeability and
“back-box” nature of AI.
Statement of interest: The author has no conflict of interest
to declare. Comments on limitations of AI are acknowledged to David
Epstein’s 2019 book “Range”.