Editor: In the following we would like to respond to the e-letters
referring to our short report:
1.) Excluding all caesarean sections instead of elective caesarean
sections only (James King)
The database we used contains an item which indicates whether a
primary (elective) section or a secondary (emergency) section was
performed. However, due to previous validation of the database we knew
that the distinction between primary and secondary caesarean section was
interpreted differently by the obstetricians documenting the births.
We therefore used a straightforward approach and excluded all
caesarean sections.
2.) Mothers fatigue may be also important (Donald Fraser Miles)
Interestingly the comments given by Donald Fraser Miles seem very
similar to those written by medical professionals when commenting on
alternative explanations for the circadian mortality gradient. Therefore
our comments given to alternative explanations towards the end of our
reply may also serve as answers to the comments by Mr. Miles. The main
difference to the comments of medical professionals seems to be that Mr.
Miles also considers induction of births as an alternative strategy to
reduce excess neonatal mortality at night. At this point we would simply
like to add that the discussion of planing the timing of birth (programmed
birth) is not new in obstetrics (1,2). We agree with Donald Fraser Miles
that this alternative would need careful evaluation before making use of
it.
3.) Night time fetal death and maternal sleep (Jason Collins)
We appreciate the perinatal physiologic speculations of Dr. Collins
as well as the reference to his own scientific work (3). Unfortunately
almost all of his remarks refer to "fetal death" which was not the subject
of our analyses. As noted in our paper we did not analyse stillbirths but
deaths during labour and in the first week after birth. We therefore
excluded all infants deceased before birth, all deaths for whom the time
of death was not known as well as those who died before arrival to the
birth clinic. We have performed additional analyses with time of death as
predictor, or stillbirths as endpoint without observing a circadian
mortality gradient within our data so far. These results may be seen as an
argument against Dr. Collins hypotheses that reduced maternal blood
pressure resulting in births of previously compromised fetuses can mainly
be assumed to have caused early neonatal deaths analysed for this article.
Additionally it is not clear how the mechanism proposed by Dr.
Collins may be an explanation for the increased weekend mortality he
refers to.
4.) Other causes predisposing to increased night-time perinatal
mortality have not been excluded (Rajesh Varma et al.)
In their first point they argue a comprehensive index case record
interrogation was absent in our data and describe in the following the
advantages of the Confidential Enquiry into Stillbirths and Deaths in
Infancy (CESDI) (4). Like presumably anyone else we agree that the CESDI
has several serious advantages compared with our data. Nevertheless our
data/ results indicate an excess early neonatal mortality at night. Varma
et. al.'s first point contains no serious argument why our results should
be biased in any specific direction.
Despite this fact Varma et al. try to imply they had already shown
our results were biased by referring to our selection criteria in their
second point ("further bias arose"). We would appreciate it if further
discussion by Varma et al. or by others referring to this important topic
would not rely on such merely rhetoric arguments.
Varma et al. criticise that our selection criteria are too imprecise.
They are however much more precise than the ones in the British studies we
referred to (5,6) and were even more precise in earlier drafts of our
paper. We therefore wonder why Varma et al. failed to comment on these
reports published just two years ago, especially as the results of those
studies seem to be much more relevant to them: Both refer to British
populations and to a health system they are currently working in.
Furthermore Varma et al. complain our early neonatal death rate is
too low which means nothing more that our selection criteria is too
rigorous. In our article we had however explained why we applied the
selection criteria used in our analysis: "There was some concern, that
babies born during daytime are more likely preterm or high risk babies who
have had induced births. We therefore restricted our study population
rigorously". In fact it is an easy task to demonstrate that this is the
case. Therefore our selection criteria is more likely to avoid bias than
to produce it. Unfortunately again Varma et al. failed to follow this
important argument when commenting our article.
In their third point Varma et al. repeat an argument they had already
presented within their first point, this time referring especially to the
diagnosis "asphyxia related deaths".
Hereby it is important to mention that we had already critically discussed
exactly this topic in our short report, but we added: "Nevertheless a
higher early-neonatal mortality in general and a higher mortality related
to asphyxia was seen". Once more it is unclear why Varma et al. are
unwilling to take our discussion of this topic into account, when
commenting our article.
In the same paragraph the authors state that not all hypoxia during
birth were preventable.
We agree and would like to add that further assumptions have to be made
before it can be assumed that excess deaths at night are not or less
preventable. Additionally we would like to note that Varma et al. give no
conclusive hint why hypoxia should be less preventable at night than
during daytime.
In the following paragraph Varma et al. offer several advises how our
article could have been improved. First they refer again to the
classification of death and recommend reclassification into the amended
Whigglesworth classification (7,8). Unfortunately Varma et al. do not seem
to know our data-set as well as we knew the Whigglesworth classification
when writing our paper. In fact we have tried to perform exactly this
reclassification but had to realise that this would produce very critical
results as the classifications used are too different. As the diagnosis
was additionally based only on the obstetrician in charge of the birth we
decided not to focus our analyses so much on the diagnosis given and more
on the total numbers of death during the early neonatal period.
Varma et al. recommend usage of several additional methods, which all
aim at ruling out or assessing the effect of other explanatory factors
than decreased medical care at night on the observed early neonatal
mortality gradient, e.g. controlling for the hypothetical fact that excess
neonatal deaths were to a substantial amount result of a higher frequency
of "high risk pregnancies at night". As several other comments refer to
almost identical topic, we would like to comment on them together:
James Collins suggested a lowered maternal blood pressure during the
night in previously compromised fetuses as a possible causal factor,
increased pregnancy risk of night births have been previously discussed
e.g. by Murphy (9) or Ruffieux et al. 1992 (10).We agree that this is an
important topic. In the letter accompanying the first draft of our paper
we wrote that we submitted the paper as a short report as we expected the
interest of the BMJ in German data confirming previous British analyses
would not be sufficient to give us room for a normal paper. We added that
we were willing to provide additional information/ analyses if requested
and have tried to do so in the following first revision of our paper but
were then advised to cut it out in order not to exceed the word limit,
which we were told to be interpreted in a strict sense. It is common sense
that a paper of a maximum of 600 words necessarily leaves many questions
open. (The comments on our paper currently count 917 words). Since the
submission of the first draft of our paper in January 2000, we have
nevertheless extended our database, performed additional detailed analyses
taking alternative risk factors of nocturnal early neonatal mortality into
account as well as written a new article, which we will submit within the
next weeks.
Other meanwhile published data by Kumar
(http://www.bmj.com/cgi/eletters/321/7256/274) confirmed a circadian
gradient of early neonatal mortality without detecting remarkable
differences in maternal complications predisposing for perinatal asphyxia.
Additionally, the response of Westgate and Gunn may be interpreted as
pointing into a similar direction.
It has to be kept in mind, however, that we are dealing with
observational studies. No evidence in the strict meaning of the word can
be expected from these kind of studies. Even after performing additional
and more detailed analyses it is common sense that alternative not
necessarily mutually exclusive explanations (11) will still be discussed
or cannot be completely ruled out.
The only alternative to provide such evidence would be conducting a
randomised clinical trial. However, such a trial would be very difficult
to perform. Very large numbers would have to be included and even more
important a randomisation into a high and low care group will be very
likely to be regarded as unethical. Maybe if Varma et al. were members of
an ethical committee would decide that such a trial is ethical to perform;
we would not.
Therefore, knowing the results of our currently unpublished analyses,
we believe it is justified to consider "better designed shifts, resulting
in shorter working hours or decreased workload with greater supervision by
experienced staff at night (...) to reduce early neonatal mortality during
the night" (12).
Yours sincerely
Günther Heller, MD
Björn Misselwitz, MD
Stephan Schmidt, MD
References:
(1) Wulf KH. Die programmierte Geburt. Therapeutische Umschau
1981;38:1009-1014.
(2) Maslow AS, Sweeney AL. Elective induction of labor as a risk
factor for cesarean delivery among low-risk women at term. Obstet Gynecol
2000;95:917-922
(3) Collins JH. Umbilical cord accidents--time of death? Am J Obstet
Gynecol 1997;177: 1566
(4) Confidential Enquiry into Stillbirths and Deaths in Infancy.
Sixth Annual Report. London: Maternal and Child Health Research
Consortium.1999
(5) Stewart JH, Andrews J, Cartlidge PHT. Numbers of deaths related
to intrapartum asphyxia and timing of birth in all Wales perinatal survey.
1993-1995. BMJ 1998;316:657-660
(6) Chalmers JWT, Shanks E, Paterson S, McInneny K, Baird D, Penney
G. Scottish data on intrapartum related deaths are in same direction as
Welsh data. BMJ 1998;317:539
(7) Keeling JW, MacGillivray I, Golding J, Wigglesworth J, Berry J,
Dunn
PM. Classification of perinatal death. Arch Dis Child 1989;64:1345-1351
(8) Patrick H T Cartlidge, Andrew T Dawson, Jane H Stewart, Gordan M
Vujanic.
Value and quality of perinatal and infant postmortem examinations: cohort
analysis of 400 consecutive deaths. BMJ 1995;310:155-158
(9) Murphy DJ. Denominators are needed before conclusions can be
drawn. BMJ 1998;316:1318
(10) Ruffieux C, Marazzi A, Paccaud F. The circadian rhythm of the
perinatal mortality rate in Switzerland. AJE 1992;135:936-952
(11) Macfarlane A. Variation in number of births and perinatal
mortality by day of week in England and Wales. BMJ 1978(2):1670-1673
(12) Heller G, Misselwitz B, Schmidt S. Early neonatal mortality,
asphyxia related deaths, and timing of low risk births in Hesse, Germany,
1990-8: observational study. BMJ 2000;321:274-275
Rapid Response:
Reply to e-letters 1 to 4
Editor: In the following we would like to respond to the e-letters
referring to our short report:
1.) Excluding all caesarean sections instead of elective caesarean
sections only (James King)
The database we used contains an item which indicates whether a
primary (elective) section or a secondary (emergency) section was
performed. However, due to previous validation of the database we knew
that the distinction between primary and secondary caesarean section was
interpreted differently by the obstetricians documenting the births.
We therefore used a straightforward approach and excluded all
caesarean sections.
2.) Mothers fatigue may be also important (Donald Fraser Miles)
Interestingly the comments given by Donald Fraser Miles seem very
similar to those written by medical professionals when commenting on
alternative explanations for the circadian mortality gradient. Therefore
our comments given to alternative explanations towards the end of our
reply may also serve as answers to the comments by Mr. Miles. The main
difference to the comments of medical professionals seems to be that Mr.
Miles also considers induction of births as an alternative strategy to
reduce excess neonatal mortality at night. At this point we would simply
like to add that the discussion of planing the timing of birth (programmed
birth) is not new in obstetrics (1,2). We agree with Donald Fraser Miles
that this alternative would need careful evaluation before making use of
it.
3.) Night time fetal death and maternal sleep (Jason Collins)
We appreciate the perinatal physiologic speculations of Dr. Collins
as well as the reference to his own scientific work (3). Unfortunately
almost all of his remarks refer to "fetal death" which was not the subject
of our analyses. As noted in our paper we did not analyse stillbirths but
deaths during labour and in the first week after birth. We therefore
excluded all infants deceased before birth, all deaths for whom the time
of death was not known as well as those who died before arrival to the
birth clinic. We have performed additional analyses with time of death as
predictor, or stillbirths as endpoint without observing a circadian
mortality gradient within our data so far. These results may be seen as an
argument against Dr. Collins hypotheses that reduced maternal blood
pressure resulting in births of previously compromised fetuses can mainly
be assumed to have caused early neonatal deaths analysed for this article.
Additionally it is not clear how the mechanism proposed by Dr.
Collins may be an explanation for the increased weekend mortality he
refers to.
4.) Other causes predisposing to increased night-time perinatal
mortality have not been excluded (Rajesh Varma et al.)
In their first point they argue a comprehensive index case record
interrogation was absent in our data and describe in the following the
advantages of the Confidential Enquiry into Stillbirths and Deaths in
Infancy (CESDI) (4). Like presumably anyone else we agree that the CESDI
has several serious advantages compared with our data. Nevertheless our
data/ results indicate an excess early neonatal mortality at night. Varma
et. al.'s first point contains no serious argument why our results should
be biased in any specific direction.
Despite this fact Varma et al. try to imply they had already shown
our results were biased by referring to our selection criteria in their
second point ("further bias arose"). We would appreciate it if further
discussion by Varma et al. or by others referring to this important topic
would not rely on such merely rhetoric arguments.
Varma et al. criticise that our selection criteria are too imprecise.
They are however much more precise than the ones in the British studies we
referred to (5,6) and were even more precise in earlier drafts of our
paper. We therefore wonder why Varma et al. failed to comment on these
reports published just two years ago, especially as the results of those
studies seem to be much more relevant to them: Both refer to British
populations and to a health system they are currently working in.
Furthermore Varma et al. complain our early neonatal death rate is
too low which means nothing more that our selection criteria is too
rigorous. In our article we had however explained why we applied the
selection criteria used in our analysis: "There was some concern, that
babies born during daytime are more likely preterm or high risk babies who
have had induced births. We therefore restricted our study population
rigorously". In fact it is an easy task to demonstrate that this is the
case. Therefore our selection criteria is more likely to avoid bias than
to produce it. Unfortunately again Varma et al. failed to follow this
important argument when commenting our article.
In their third point Varma et al. repeat an argument they had already
presented within their first point, this time referring especially to the
diagnosis "asphyxia related deaths".
Hereby it is important to mention that we had already critically discussed
exactly this topic in our short report, but we added: "Nevertheless a
higher early-neonatal mortality in general and a higher mortality related
to asphyxia was seen". Once more it is unclear why Varma et al. are
unwilling to take our discussion of this topic into account, when
commenting our article.
In the same paragraph the authors state that not all hypoxia during
birth were preventable.
We agree and would like to add that further assumptions have to be made
before it can be assumed that excess deaths at night are not or less
preventable. Additionally we would like to note that Varma et al. give no
conclusive hint why hypoxia should be less preventable at night than
during daytime.
In the following paragraph Varma et al. offer several advises how our
article could have been improved. First they refer again to the
classification of death and recommend reclassification into the amended
Whigglesworth classification (7,8). Unfortunately Varma et al. do not seem
to know our data-set as well as we knew the Whigglesworth classification
when writing our paper. In fact we have tried to perform exactly this
reclassification but had to realise that this would produce very critical
results as the classifications used are too different. As the diagnosis
was additionally based only on the obstetrician in charge of the birth we
decided not to focus our analyses so much on the diagnosis given and more
on the total numbers of death during the early neonatal period.
Varma et al. recommend usage of several additional methods, which all
aim at ruling out or assessing the effect of other explanatory factors
than decreased medical care at night on the observed early neonatal
mortality gradient, e.g. controlling for the hypothetical fact that excess
neonatal deaths were to a substantial amount result of a higher frequency
of "high risk pregnancies at night". As several other comments refer to
almost identical topic, we would like to comment on them together:
James Collins suggested a lowered maternal blood pressure during the
night in previously compromised fetuses as a possible causal factor,
increased pregnancy risk of night births have been previously discussed
e.g. by Murphy (9) or Ruffieux et al. 1992 (10).We agree that this is an
important topic. In the letter accompanying the first draft of our paper
we wrote that we submitted the paper as a short report as we expected the
interest of the BMJ in German data confirming previous British analyses
would not be sufficient to give us room for a normal paper. We added that
we were willing to provide additional information/ analyses if requested
and have tried to do so in the following first revision of our paper but
were then advised to cut it out in order not to exceed the word limit,
which we were told to be interpreted in a strict sense. It is common sense
that a paper of a maximum of 600 words necessarily leaves many questions
open. (The comments on our paper currently count 917 words). Since the
submission of the first draft of our paper in January 2000, we have
nevertheless extended our database, performed additional detailed analyses
taking alternative risk factors of nocturnal early neonatal mortality into
account as well as written a new article, which we will submit within the
next weeks.
Other meanwhile published data by Kumar
(http://www.bmj.com/cgi/eletters/321/7256/274) confirmed a circadian
gradient of early neonatal mortality without detecting remarkable
differences in maternal complications predisposing for perinatal asphyxia.
Additionally, the response of Westgate and Gunn may be interpreted as
pointing into a similar direction.
It has to be kept in mind, however, that we are dealing with
observational studies. No evidence in the strict meaning of the word can
be expected from these kind of studies. Even after performing additional
and more detailed analyses it is common sense that alternative not
necessarily mutually exclusive explanations (11) will still be discussed
or cannot be completely ruled out.
The only alternative to provide such evidence would be conducting a
randomised clinical trial. However, such a trial would be very difficult
to perform. Very large numbers would have to be included and even more
important a randomisation into a high and low care group will be very
likely to be regarded as unethical. Maybe if Varma et al. were members of
an ethical committee would decide that such a trial is ethical to perform;
we would not.
Therefore, knowing the results of our currently unpublished analyses,
we believe it is justified to consider "better designed shifts, resulting
in shorter working hours or decreased workload with greater supervision by
experienced staff at night (...) to reduce early neonatal mortality during
the night" (12).
Yours sincerely
Günther Heller, MD
Björn Misselwitz, MD
Stephan Schmidt, MD
References:
(1) Wulf KH. Die programmierte Geburt. Therapeutische Umschau
1981;38:1009-1014.
(2) Maslow AS, Sweeney AL. Elective induction of labor as a risk
factor for cesarean delivery among low-risk women at term. Obstet Gynecol
2000;95:917-922
(3) Collins JH. Umbilical cord accidents--time of death? Am J Obstet
Gynecol 1997;177: 1566
(4) Confidential Enquiry into Stillbirths and Deaths in Infancy.
Sixth Annual Report. London: Maternal and Child Health Research
Consortium.1999
(5) Stewart JH, Andrews J, Cartlidge PHT. Numbers of deaths related
to intrapartum asphyxia and timing of birth in all Wales perinatal survey.
1993-1995. BMJ 1998;316:657-660
(6) Chalmers JWT, Shanks E, Paterson S, McInneny K, Baird D, Penney
G. Scottish data on intrapartum related deaths are in same direction as
Welsh data. BMJ 1998;317:539
(7) Keeling JW, MacGillivray I, Golding J, Wigglesworth J, Berry J,
Dunn
PM. Classification of perinatal death. Arch Dis Child 1989;64:1345-1351
(8) Patrick H T Cartlidge, Andrew T Dawson, Jane H Stewart, Gordan M
Vujanic.
Value and quality of perinatal and infant postmortem examinations: cohort
analysis of 400 consecutive deaths. BMJ 1995;310:155-158
(9) Murphy DJ. Denominators are needed before conclusions can be
drawn. BMJ 1998;316:1318
(10) Ruffieux C, Marazzi A, Paccaud F. The circadian rhythm of the
perinatal mortality rate in Switzerland. AJE 1992;135:936-952
(11) Macfarlane A. Variation in number of births and perinatal
mortality by day of week in England and Wales. BMJ 1978(2):1670-1673
(12) Heller G, Misselwitz B, Schmidt S. Early neonatal mortality,
asphyxia related deaths, and timing of low risk births in Hesse, Germany,
1990-8: observational study. BMJ 2000;321:274-275
Competing interests: No competing interests