J. Physiol. (1984), 347, pp.

1-16
With 8 text-figure8
Printed in Great Britain

THE DEVELOPMENT OF STABILITY OF RESPIRATION IN HUMAN

INFANTS: CHANGES IN VENTILATORY RESPONSES TO
SPONTANEOUS SIGHS

BY PETER J. FLEMING, ARTHUR L. GONCALVES, MICHAEL R. LEVINE

AND SUSAN WOOLLARD
From the Departments of Child Health and Physiology, Bristol University
(Received 14 December 1982)
SUMMARY

1. Serial respiratory recordings using impedance pneumography and barometric

plethysmography were made from shortly after birth to 7 months in fifteen normal
full-term infants. Each recording was made with the infant asleep and sleep state was
estimated from records of electroencephalogram and electro-oculogram made in
parallel.
2. The respiratory records obtained during non-rapid eye movement (r.e.m.) sleep
were analysed with computer assistance and stretches of the record, approximately
1 min before and up to 2 min after a spontaneous sigh and ensuing apnoeic pause,
were processed and presented as sequential values of the fractional deviation of '.
the breath by breath minute volume, from the mean. That part of the sequence which
represented the respiratory response to the sigh was then fitted with second order
equations representing the critically or underdamped response. The results were
presented for each curve in terms of 6, the damping ratio and Wn, the frequency of
the undamped respiratory oscillation.
3. Three-quarters of the responses could be so fitted with an error of 20 % or less.
The residual responses were mainly from infants within a few days of birth. In the
youngest infants (4 days or less), the respiratory response to a sigh was highly stable
but sluggish: during the period 4-8 days to 3-4 months, the oscillatory period
diminished from ca. 25-12 s and respiration was potentially unstable since a small
reduction in the damping factor would cause prolonged oscillation while, from 3-4
months, the more mature type of response which was stable with a rapid recovery
supervened.
4. The possible mechanisms responsible for this trend are discussed in terms of the
factors thought to determine respiratory stability in the adult together with the
possible relevance of the results to the normal process of respiratory adaptation at
birth and to the respiratory difficulties encountered by some infants in the new-born
period and early infancy.
INTRODUCTION

Shortly after birth, breathing becomes continuous and, after some hesitation,
regular but there is some suggestion that full respiratory control is not achieved
1

PHY 347

P. J. FLEMING AND OTHERS

until days or even weeks after birth. The mature respiratory response to hypoxia
consisting of a sustained hyperpnoea is not seen until the end of the first week in the
human infant (Brady & Ceruti, 1966) or until 6 weeks in the kitten (Schweiler, 1968).
The sensitivity of the respiratory response to inhaled CO2 whether measured under
steady-state conditions (Rigatto, Brady & Torre Verduzco, 1975; Krauss, Klain,
Waldman & Auld, 1975; Frantz, Adler, Thach & Taeusch, 1976), or as the pressure
generated by occlusion at end-expiration (Thach & Taeusch, 1976) is reported to
increase with post-natal age while it is only from about the 20th day after birth that
the respiratory response to CO2 is more marked in non-rapid eye movement (r.e.m.)
sleep than in r.e.m. sleep in the human infant (Davi, Sankaran, McCallum, Cates &
Rigatto, 1979) and in the monkey (Guthrie, Standaert, Hodson & Woodrum, 1980)
as it is reported to be in the adult (Phillipson, Murphy & Kozar, 1976). In addition,
periodic respiration commonly occurs in a substantial proportion of normal, full-term
infants for up to 6 weeks or more (Hoppenbrouwers, Hodgeman, Harper, Hoffman,
Sterman & McGinty, 1977).
This last observation raises the question as to how stable respiration actually is
in the new-born period and whether in view of the changes in respiratory chemoreflex
activity, outlined above, there is a period of uncertainty in respiratory control. Such
a period of uncertainty would not be unexpected in a control system which had not,
prior to birth, had to make accurate responses to chemical, thermal, metabolic and
other stimuli while the interest of such a period, if it could be shown, would be that
it might explain the origin and nature of the respiratory difficulties encountered by
a proportion of infants in the new-born period. This point has not yet been studied
systematically. Hathorn (1978), using cross-correlation techniques, showed that the
majority of infants in a small series in the first week after birth were able to maintain
changes in tidal volume and respiratory frequency exactly out of phase, an
arrangement which should make for a stable respiratory output. But whether other
factors thought to influence respiratory stability such as lung-to-receptor circulation
time, CO2 and 02 controller gain and tissue stores also operate in the new-born and
whether such factors alter with post-natal age, is not known.
We have approached this question by analyzing the respiratory response to
spontaneous sighs and the ensuing apnoeic pause (Cross, 1954) in a group of normal,
full-term infants serially from birth until 7 months of age and have estimated
respiratory stability at each age in terms of the damping coefficient and oscillatory
period. This has shown a clear trend with post-natal age and the results are discussed
in terms of the normal respiratory adaptation at birth and in deviations from this
which might affect respiratory control. A preliminary report of this work has been
published as an abstract (Fleming & Levine, 1982).
2

METHODS

Fifteen, normal, full-term infants were studied with the informed consent of their mothers, many
of whom assisted at the recording sessions. Three mothers received pethidine more than 6 h before
birth; one mother received pethidine within 2 h of birth while the others received no analgesia or
NO2 and 02 only. All were vaginal-vertex deliveries and none required resuscitation. Respiratory
recordings were made with the infants asleep, most commonly after a feed. Recordings lasted for
as long as possible: in one infant study only 6 min of recording was possible but in all the other

VENTILATOR Y RESPONSES TO SIGHS IN INFANTS

tests, recording lasted for 20 min or longer. In all infants, respiration was measured in terms of
the output of a 4-lead, transthoracic impedance pneumograph (Olsson & Victorin, 1970) and in
twelve infants, in addition, by means of a barometric plethysomograph based on the technique
described by Drorbough & Fenn (1955) and which has been shown to give accurate and consistent
results for the measurement of tidal volume and respiratory frequency (Epstein, Epstein, Haddad
& Mellins, 1980). The sleep state of the infants was estimated from parallel recordings of the
electroencephalogram (e.e.g.) and electro-oculogram (e.o.g.) using bilateral electrodes and the
observation of eye movements (Anders, Emde & Parmalee, 1971).
In preliminary tests, a close correlation was demonstrated between the outputs of the impedance
pneumograph and that of a pneumotachograph, positioned in a face mask. This served to calibrate
the pneumograph tracing but in the same series of tests, the results of which have been published
elsewhere (Fleming, Levine & Goncalves, 1982), it was also shown that application of the face mask
significantly distorted the breathing pattern. These tests also showed that the pneumograph output
was affected by non-specific movements of the body but that these could be distinguished from
spontaneous sighs by recording tidal C02, sampled from the nose at a flow rate of 100 ml minand analysed with a Beckman LB2 infra-red analyzer. The outputs of e.e.g., e.o.g. and pneumograph
electrodes, the pressure transducer within the plethysmograph and the CO. analyzer were recorded
in parallel on a multi-channel recorder (Devices M19) while the respiratory signals were also
recorded on FM tape (Racal-Thermionic, Store-4 recorder).

Anoly8i8 of respiratory data

The respiratory data were subsequently sampled at either 20 or 32 ms intervals using ADC (ARII)
of a PDP 11/10 laboratory computer, and the sampled signals stored on disc. These signals were
first used to plot a respiratory trace on tracing paper which could be superimposed on the analog
plot of the pneumograph or plethysmograph in each test so that (a) the sampling routine could
be verified and, (b) from the parallel e.e.g. and e.o.g. signals, the periods of r.e.m. and non-r.e.m.
sleep could be identified. Next, the sampled signals on disk were processed using a peak detection
program to yield sequential points of maximum voltage deviation corresponding to peak inspiration
and expiration. In order to identify these points with certainty in the presence of artifacts such
as the heart beat, two parameters were used which could be varied at the time when the program
was run. These were: the minimum time and the minimum voltage change between end-inspiration
and end-expiration or vice versa. Their value was determined from inspection of the respiration
trace plotted by computer and if they were correct, this meant that the program did not identify
as breaths any excursions in the recorded signal which were smaller than the window defined by
these values.
The output of the peak detection program was also plotted on tracing paper as a bar chart as
is shown in Fig. 1. Each breath is represented by two vertical and two horizontal lines which are
proportional to, respectively, inspiratory and expiratory volume (VTj, VTe) and duration (T1, Te).
This bar chart was superimposed on the computer plot of the respiratory trace and if more than
one breath in 3 min was wrongly identified by the program, the value ofthe parameters was changed
and the program re-run. A number of respiratory variables, such as the ratio of tidal volume to
inspiratory duration, total breath duration and minute volume of ventilation, could be derived from
these primary variables, which were stored on disk, but, in the present study, only sequential breath
by breath values for tidal volume (VT), respiratory frequency (f) and their product, the breath by
breath minute volume of ventilation (PGE), were presented. Each of these variables was plotted on
the ordinate against time on the abscissa and, in order to facilitate comparison between the results
in different tests, all values were expressed and plotted as the fractional differences from the mean
of that study, (y-y)/y. This readily allowed the pattern of change of each variable after any
disturbance to be visualized, Fig. 1.
For the purpose of the present study, only the respiratory responses to a spontaneous sigh and
the apnoeic pause which followed it were analysed. A sigh was defined as a breath which was at
least twice as large as the other breaths in the sequence and which -was followed by a transient
fall in ventilation or an apnoeic pause which was itself twice as long as the average breath duration.
Further, because of the irregularity of breathing in r.e.m. sleep, only the responses in non-r.e.m.
sleep were analysed. Typically, as is shown in Fig. 1, a stretch of breathing recorded approximately
1 min before and up to 2 min after each sigh was selected for analysis. As will be shown in the Results

section, there was considerable variation in the respiratory responses at various post-natal ages,
1-2

P. J. FLEMING AND OTHERS

4
A

AnAAAAAftitRl

if/<A

. ArtrAAllt\4A MltAAAAr

,AA14AAA444AA4

Vi

L_

Insp.

_ _ _ _ I_

ir

F 1 _ ~~n

VO

dL

P2i Q
20 s

Fig. 1. Traces to show the sequence of processing. A, the computer plot of the output of
the impedance pneumogram (inspiration (insp.) downwards and in arbitrary units)
including a spontaneous sigh and apnoeic pause. B, an enlargement of the trace within
the rectangle in A with super-imposition of the output ofthe peak detection program where
the vertical line p is proportional to inspiratory tidal volume (VTi), q to inspiratory time
(Ti), r to expiratory tidal volume (VTe) and 8 to expiratory time (Te). C, is the output of
the peak detection program corresponding to the whole of trace A. D, E and F show
sequential values, corresponding to the breaths in A of the fractional difference of,
respectively, f, VT and PE from the mean on the ordinate against time on the abscissa.
Where
V = (VTi + VTe) f= T + TX and PE = VT Xf.
In F, the respiratory response has been fitted with a damped cosine curve. Traces A and
C-F inclusive are shown on the same time scale.

but there was also variation in the separate responses of VT andf. On some occasions, both VT and
f were initially depressed following the disturbance and then recovered monotonically to control
levels. On other occasions, VT responded with an oscillatory pattern whilef responded monotonically.
On yet other occasions, both VT andf responded with an oscillatory pattern. We judged that values
for PE, the minute volume of ventilation, was the simplest and most consistent estimate of
respiratory output and in the present study, respiratory responses have been analysed in terms
of this variable only.
The simplest type of system which responds to a transient disturbance in a similar way to the
monotonic and oscillatory respiratory patterns which we observed may be described by a second
order, differential equation:
d2y/dt2 + 2un gdy/dt + wn2y = F(t),
(1)
where y = (PE -fE)/ TE and where f is the damping factor and Wn is the natural frequency of
oscillation (radians s-1) (Burghes & Graham, 1980).

VENTILATORY RESPONSES TO SIGHS IN INFANTS

Here F(t) is a function defining the imposed transient disturbance due to the sigh. Zero time is
measured from the start of recovery of the sigh. It is assumed that F(t) = 0 for t > 0. The two
constants C and Cn determine the characteristics of the system (Bayliss, 1966). We can classify the
types of response to a transient disturbance in terms of the value of C. If C < 0 then the response
takes the form of an increasing oscillation and the system is unstable. If 6 = 0 then the system is
undamped and will oscillate about the equilibrium state continuously with an angular frequency
Wn. If 0 < C < 1 then the system is underdamped and will oscillate with decreasing amplitude. If
C = 1 then the system is critically damped and will return to equilibrium in the shortest possible
time without oscillating. If C > 1 then the system is over-damped and will return asymptotically
to equilibrium, again without oscillating. We used two classes of solution of t (for t > 0): First, for
critical damping ( = 1),

(2)
y = yoe wnt (1 + Wn ),
here yo is the amplitude of the initial deviation.
Secondly for the underdamped response (C < 1),
y = Aef'wntcos[(nV(1 ) .t+q$],
(3)
here A is the amplitude of the oscillation obtained when there is no damping, i.e. for =00 and
9 has the form of a phase factor. yo, A,9 are constants whose values are dependent upon the precise
value of F(t) and since this is undefined by our measurements we cannot assign any significance
to them. The effect of F(t) is to determine the value of y and its derivative dy/dt at t = 0. We have
assumed that for the early responses which appeared to fit the critically damped case that the
derivative of y at time t = 0 is zero; which gives eqn. (2) in which there is one undetermined
constant. This assumption was not made for the underdamped case and so we have two
undetermined constants A and 0. The value of 0 can be estimated by inspection of the ventilation
(Fig. 7). At the start of the recovery after the sigh (i.e. at t = 0) y is negative and rising. The
argument of the cosine must be between 2fr and 3fr at t = 0, which means that 2fr < 0 < 3X7. It
is easy to show that the ratios of successive maxima are independent of A and q$ as are the intervals
between successive zero crossings.
Curve fitting procedure
Each respiratory response to a sigh was first visually inspected and a decision made as to whether
it could or could not be fitted by either of the eqns. (2) or (3). If the curve was thought capable
of being fitted, the appropriate equation was used and approximate values inserted: (yo, an for eqn.
(2); A, fi, T, 0 for eqn. (3) where fi = - f and T = 2ff/conV 1 -i2). The calculated curve was
superimposed on the original response curve and the parameters adjusted by trial and error until
an approximate match was obtained. The parameters were then refined by an iterative procedure
which minimized the mean sum of squares of the differences between observed and calculated curve,
E, where
E=N

ob

In this procedure (Kowalik & Osborne, 1968) the parameters were altered by fixed amounts one
at a time. For each new parameter the alteration in the residual E was determined. That new
parameter which gave the largest fall in E was adopted and the procedure repeated. When none
of the shifted parameters gave a fall in E then the shifts themselves were halved one at a time to
find which new shift gave the largest fall in E. This new shift was adopted and the minimization
was continued in this way until the changes in E and the fitted parameters were small. This
procedure converged rapidly. Finally the fitted curve was plotted superimposed upon the data. In
order to compare the fit of responses of different magnitude, E/A max was calculated for each fitted
curve where A max is the maximum displacement of y following a sigh.
RESULTS

The ages at which recordings were made in individual infants are shown for the
series as a whole in Fig. 2. Recordings were made in all infants in the first 1-2 days:

P. J. FLEMING AND OTHERS

Age of babies at time of recording 0

Baby
number
4 _
R.
U l

*e

IIU.

12
13
19
20
21.
22

?'41
;-h

261

..
8*

2)'7/ l0

on

z71

2
34
Days

3844926
B 51 68 41

810
Age

12

14

16

18 20 2224 26
Weeks

28

33

18

28

11

12

13

36

25

33

12

12

12

17

30

Fig. 2. A summary of the ages of individual infants at which respiratory recordings were
made. Below: B the total number of spontaneous sighs recorded in non-r.e.m. sleep within
each age range indicated and A the corresponding number of response curves which could
be fitted.
0*9
0-8
0-7
C

.E
-C

0-6

0-5-

en

0*4

0-2
01

57-112
113-224
13-24 25-48 49-95 96-144 15-28 29-42
Hours
Days
Fig. 3. A graph to show the relation between the frequency of sighs recorded during
non-r.e.m. sleep and post-natal age. Each point represents the mean frequency of sighs
min-' + s.E. of mean.

in thirteen infants, three or more recordings were made and in nine, the period of
observation was 4 months or more. This Figure also summarizes the total number of
sighs recorded in non-r.e.m. sleep within the ranges of post-natal age indicated (B)
together with the number of respiratory responses which could be satisfactorily fitted
(A). It is clear that many more sighs were recorded in the younger infants. This was

7
VENTILATOR Y RESPONSES TO SIGHS IN INFANTS
only in part due to the fact that more recordings were made in these infants. It was
also due to the fact, as is illustrated in Fig. 3, that the frequency of sighs diminished
with post-natal age so that from approximately 2 months onwards, only a very few
sighs occurred during a typical 20-30 min period of non-r.e.m. sleep. This paucity
was enhanced by another factor. In a proportion of older infants, respiration
recovered so rapidly to control levels following a sigh and apnoeic pause, that the
response could not be fitted satisfactorily. This point is discussed below. We also
observed that, as is illustrated in Fig. 6, with increasing post-natal age, frequency
of respiration diminished and, in general, that the frequency and depth of breathing
became much more regular.

2A

-2

-2

-2

-2

C~~~~~~~~~2
20 s

Fig. 4. Traces illustrating respiratory response curves which could not be satisfactorily
fitted with cosines. For description, see text. Each trace is presented as the fractional
difference of values of PE from the mean.

For a number of reasons, a proportion of respiratory responses to spontaneous sighs

could not be fitted with either eqn. (2) or (3). First, there were sighs or augmented
breaths which were not followed by apnoeic pauses or periods of hypoventilation and
which were followed by no discernible respiratory response. Such transients, illustrated
in Fig. 4A, were ignored and no attempt was made to fit them. Secondly, in twelve
cases, a spontaneous sigh and ensuing apnoeic pause was followed by a second sigh
with varying latency (Fig. 4B), the subsequent respiratory response was similarly
unfittable. Thirdly, in eight cases, respiration returned so rapidly to control levels

P. J. FLEMING AND OTHERS

following a sigh that all that could be said was that the responses were very highly
damped. Three of these were in infants of less than 3 days of age: the remaining five
were in infants of 3 months or more and in these, fitting was made yet more difficult
because the relatively low frequency of respiration yielded very few points for fitting.
Lastly, in fifty-one cases, the respiratory responses to spontaneous sighs and apnoeic
pauses were so irregular, as is shown in Fig. 4C, or so slight as to be indistinguishable
from noise that fitting was either impossible or possible but only with a very large
error.
,. 222018 -AA

i 16 14 -AAAA
12 -

AAAAAAA

~~A A

CU
0

M
64- 6
A

0
Ar

&&AA
AA

OA

000

1000 U

aoooo
00 0

0ooo0o9ooooooooooooogo"
0-05

0-10

0-15

0-20

ol IONI 1m51a

l
0-25 0-30
E/A max

0-35

0-40

0-45

Eo 05 1

0-50

0-55

Fig. 5. The distribution of E/A max (see text). Open circles represent respiratory response
curves recorded from infants less than 48 h of age: open triangles, infants of more than
48 h of age.

Together, this group of seventy-one unfittable responses accounted for slightly less
than a quarter of the total number recorded but it was unevenly distributed. It
accounted for one-third of the responses recorded in the first 8 days and for only 17 %
of the responses recorded thereafter.
In addition to those responses which could not be fitted, a number of responses
could only be fitted with rather large errors. This is illustrated in Fig. 5 in which the
distribution of E/A max has been plotted. Out of a total of 250 fitted curves, sixteen
were fitted with an error of 31 % or more and of these, all but three occurred in infants
of less than 1 day old. If an error of 20 %, which might be regarded as more acceptable,
is taken, this accounted for 188 or 75 % of the curves and of the residual sixty-two
curves with a greater error, forty-four (71 %) were found in traces from infants of up
to 48 h. For the purposes of presentation of results in this paper, e.g. Fig. 8, only
the respiratory response curves fitted with an error of 20 % or less have been
considered but the trend observed with data derived from curves fitted with an error
of 30 % or less was found to be virtually identical.
The pattern of responses to a sigh and apnoeic pause varied with post-natal age
and the sequence is illustrated by traces from one baby (26) in Fig. 6. Impedance
pneumograms before and following a sigh and apnoeic pause are shown at the ages
indicated. The same traces, processed and expressed as the fractional deviation of
P from the mean on the same time base, are shown in Fig. 7 A with the fitted response

9
VENTILATOR Y RESPONSES TO SIGHS IN INFANTS
curves superimposed. As is indicated from the values of E/A max in no case was the
fitted error greater than 15%. The values for the two constants, fi the damping
coefficient, (= - on), and T, the observed period of oscillation [= 2rr/wn v (1 - 2)]
are also given for each curve. These indicate that: at the earliest age, the response
is critically damped: at 47 h, a small overshoot appears with an oscillatory period
of 22-5 s and, with increasing post-natal age, the response becomes progressively
underdamped and the oscillatory period diminishes but, at a post-natal age of 5
months or so, the damping coefficient increases towards initial levels while T remains
unchanged or diminishes further.
+ Insp.

Age

23 h

20s

_AM4'A

VW 38 days

_Wvw

C _

47 h

94 days

fVfV\ 151 days

VVYYV_206 days

F
Sigh

Fig. 6. A series of traces, A-F, from one infant (no. 26) showing pneumograms (inspiration
(insp.) downwards) obtained at the ages indicated. Each trace includes a spontaneous sigh
and ensuing apnoeic pause. All are shown with the same time scale.

The changes in respiratory stability with post-natal age can be compactly

represented in the form of an S-plane diagram with axes , and 2ff/T, equivalent to
polar coordinates, Wn and 0. The values for , and T, derived from the curves a-f in
Fig. 7 A have been so plotted in Fig. 7 B. Since it is known that stability depends
upon both the damping factor cos 0 and the undamped frequency of oscillation Wn,
we may conclude that point a represents a highly stable but slow response; point b
a stable but faster response: points c and d relatively unstable responses while points
e andf represent a more stable response with rapid recovery. This last type of response
may be considered optimum and is consistent with the regular pattern of respiration
seen in infants in this age range. By contrast, the points c and d indicate relatively
unstable respiratory control since a small increase in the value for f such that it
became zero or positive would be associated with persistent respiratory oscillation
with maintained or increased amplitude.

P. J. FLEMING AND OTHERS

10

This trend with post-natal age was seen in all the infants in this series and the
changes in values for cos 0 and (On are summarized in Fig. 8A and B respectively.
This confirms that the respiratory responses shortly after birth were relatively highly
damped, that the degree of damping diminished over the first 6 days and thereafter
increased. Linear regression of cos 0 upon age showed a highly significant negative
trend initially (d.f. 92, t = 8-98, P < 0-001) and a less significant positive trend from
Age Sigh(S)I
a

A Curves fitted to pattern of ventilation

after a sigh

23 h
-010 T= Infinity
=3
012=(criticall

E
1

Is

amax

B Pole diagram
f
d

ly damped)

b 47 h

0-7

0*6
0-5

/2 _

fk I I

S max01=_As

0.4

Tr _ %% e, _

=zs

0-2

'

c 38 days

ES

0 -12-0

=-004

T=

14s

-0 15

Amax

2w

- a -0 0
-0. 1 0 -0-05

d 94 days

IS

Amax

=015

=-0.02

T=125s

e 151 days

Is

A9-0-094T=12p5s

0
=0-09

1.0s
01

Amax

f 206 days
=

0-10

j3-0-097 T

10-0 s

20 s

Fig. 7. A, the same traces as in Fig; 6, processed and presented in each case as the fractional
deviation of PE from the mean. Inset is the scale for ordinate and abscissa. Superimposed
on each response curve is the fitted cosine curve from which the values for the damping
factor, fi, and period of oscillation, T, given with each curve are derived together with
the error of fitting, E/A max. B, an S-plane Pole diagram in which values for fi and T
from the curves a-f in Fig. 7 A have been plotted on the axes -fl and 2v/T. An example
is given to show the derivation of On as the distance to a given point along a line from
the origin subtended by the angle 0.

the 15th day (d.f. 69, t = 1-99, P < 0 05). As is clear from the standard error of the
averaged values for cos 0 within each age range, there was considerable variation,
most marked in the youngest and oldest infants. This was due not only to the scatter
of individual values but also the variation in the time course of trends with age. Thus,
value for cos 0 in the first 24 h varied from I 0 to 0-28 and cos 0 declined at varying
rates over the subsequent days. For example, in Baby 21 cos 0 had fallen to low levels
(< 0 1) by 96 h; it then remained low for the subsequent month and had risen by
the 68th day. In Baby 26, on the other hand, cos 0 fell from 0-62 to 0 37 between

VENTILATOR Y RESPONSES TO SIGHS IN INFANTS

A

0 80-719

29

0-6-

<,0.5-Ll
0

21
7

05425
T4

03-

>

02

12

10

co

~~~~~~~I

14

LO

ey

11

26

C4
~co

"IO

.
//, .CL
0),

v' ',

Days

Hours

0-8

031-

~~~~~~~~~~~~~~~~~~~~2

0-7-

0-1-

26
~~~~~~~~~
~~6
1

06

Q414'
19 29

0-2
0.1co b
IV
e o anv
of rli
Fig
CN fitte
2
of I~
pointrSummary
CV)
O
L~0
C')
by th
24and48 itthnfellmuchmore#lowlyreachn le - hn0N 1 C 9)
Days
Hours
20t de
Fig. 8. Summary of the relation between values for A, cos 6 and B, Wn and post-natal age.
Each point represents the mean + s.E,. of the number of curves, given above each
point, fitted with an error of 20 % or less within the age ranges indicated.
0

~.

24 and 48 h; it then fell much more slowly reaching less than 0-1 by the 94th day
and was still only 0-13 on the 206th day. The curves from the nine infants followed
for 4 months or more reached their nadir between the 26th and 111th days. The trend
thereafter was much more difficult to determine because, as described above, the
number of sighs in any given recording session in the older infants was relatively few.

P. J. FLEMING AND OTHERS

12
Half of the curves showed no recovery from the lowest point: the other half showed
some recovery but in no case were the responses as highly damped as in the first days
after birth. The changes in wO, as summarized in Fig. 8B, were more straightforward.
Despite variation, wo in each infant showed a progressive rise with age: the positive
trend is significant (d.f. 172, t = 8-33, P < 0 001). Although the number of values is
small there is some suggestion above 2-3 months of age, that the frequency of the
undamped respiratory oscillation becomes constant.

DISCUSSION

The present series of experiments has involved a substantial processing chain from
raw respiratory data to the final presentation of respiratory stability in terms of
damping coefficient and oscillatory period and at each stage a number of qualifications
have to be applied which affect any conclusions which may be drawn. These are
considered in turn.
The respiratory data were obtained by non-disturbing therefore necessarily
indirect recording techniques, namely the impedance pneumograph and the barometric
plethysmograph. The most direct method, pneumotograph and face mask, introduces
changes in the pattern of breathing (Askanazi, Silverberg, Foster, Hyman, Milic-Emili
& Kinney, 1980; Fleming et al. 1982) and was used only to show that the indirect
methods were accurate. The processed data from the two outputs were virtually
identical. The disadvantages of the impedance pneumograph - its inability to
measure tidal volume quantitatively and its susceptibility to non-specific body
movements - were not found to be of any great importance since, for the purposes
of analysis, we were principally concerned with relative changes in TIE over short
periods while the use of the C02 analyser assisted in distinguishing between
respiratory and non-respiratory movements. We were therefore as certain as we could
be that the respiratory data which were subsequently processed were unaffected by
phase or other form of distortion.
This type of transient analysis could only be applied in non-r.e.m. sleep since
respiration in r.e.m. is highly irregular (but see Waggener, Frantz, Stark & Kronauer,
1982). However, it is in non-r.e.m. that control is mediated principally by inputs from
chemoreceptors and the respiratory reflexes (Bryan & Bryan, 1978).
As indicated above, difficulties were encountered in fitting a proportion of the
respiratory responses to spontaneous sighs. This was most obvious in infants in the
first few days after birth. This may simply be an index of the variability of breathing
pattern seen at this age, but not in the older infants, which has also been observed
by other workers (Hathorn, 1978; Waggener et al. 1982). The type of responses varied
considerably even in the same infant during the same recording session. It is possible
that a better description of the responses may have been achieved by the use of third
or even fourth order equations. We have not explored this possibility since our
approach in this, the first attempt to determine respiratory stability in infant or adult
in quantitative terms, has been to limit the number of constants. This possibility must
therefore remain open.

VENTILATOR Y RESPONSES TO SIGHS IN INFANTS

13

Po88ible physiological s8gniftcance

It is worth emphasizing that three-quarters of the responses could be fitted with
relatively small error and this suggests that in these cases the respiratory control
system operates, as if it were under the control of a second order feed-back loop. It
is most probable that the dominant disturbing factor to which the system was
responding was a chemical one associated with the sighs since the response lasted for
tens of seconds rather than a breath or two which would have been expected if the
response were a reflex one from, for example, lung receptors.
We have assumed that the chemical disturbance consists of a rise in Pa (Biscoe
& Purves, 1967), a rise in pHa (Band, Cameron & Semple, 1969) and a fall in Pa&co,
(Lewis, Ponte & Purves, 1980) associated with and starting 2-3 s after the sigh. This
leads to a rapid fall in carotid body chemoreceptor afferent discharge which, together
possibly with the increased discharge from slowly adapting lung receptors, causes the
period of apnoea during which time, the chemical changes in arterial blood are
reversed to or even beyond control levels. The origin of the spontaneous sigh is
uncertain. It has been proposed as representing the respiratory response to a fall in
pulmonary compliance associated with alveolar collapse and mediated by receptors
in lung and chest wall (Bendixen, Smith & Mead, 1964). If so, this may explain the
relative frequency of sighs in the immediate post-natal period before the functional
residual capacity is established (Olsson & Victorin, 1970; Thach & Taeusch, 1976).
Whatever the origin of the sigh, it has been found, in the present series, to provide
a consistent disturbance to respiration though clearly the variable depth of the sigh
and duration of the apnoeic pauses was likely to give rise to variable chemical changes
in arterial blood.
It is possible that in the first few days of life, cardiovascular responses play a major
part in the control of blood gases as in the fetus. This could explain the damped and
variable nature of the response to the sigh in the new-born.
Evidence that the blood gas disturbances due to a sigh might be smeared by an
effective right to left shunt in the first few days of life is provided by Purves (1966a)
in a study on new-born lambs. This could be due to the persistence of fetal vascular
channels. Such attenuation of the chemical changes in arterial blood may afford some
protection for the respiratory control system during the transitional period
immediately after birth when it is being faced with, and having to respond to, stimuli
never previously experienced.
By the end of the first week or so, the respiratory responses are much less variable.
By this time, the damping ratio has fallen substantially below initial levels.
Thereafter, it remains at low levels for some weeks and, in approximately half of the
infants studied, it then rose at about 3 months or more. Over the same period, the
period of the undamped respiratory oscillation approximately halved.
It is possible that it is only an increasing influence and importance of the peripheral
chemoreflex which could cause so marked a reduction in the period of the undamped
respiratory oscillation. Such an increase in the peripheral chemoreceptor effects might
arise from alterations in the signal seen by the receptor (e.g. from decreasing venous
admixture with increasing age), from increased sensitivity ofthe receptor or increased
central response to peripheral chemoreceptor inputs.

14

P. J. FLEMING AND OTHERS

The contribution of the peripheral chemoreceptors to the dynamic respiratory

response to C02 has not been widely studied but it is clear from studies in the
unanaesthetized adult dog (Dejours, 1963) and new-born lamb (Purves, 1966b) that
in the peripherally chemodenervated animal, the response has a longer latent period
and is substantially more sluggish. It would be of considerable interest to see whether
the speed of the respiratory response to a standard CO2 stimulus does alter in the
new-born period and whether such alteration could be affected by peripheral
chemodenervation.
There is similar uncertainty about the factors which determine the degree of
respiratory damping. It has been proposed in the adult that important factors include
the relative importance of tissue stores of C02 and 02, and C02 and 02 controller gains
and that, generally, C02 is the more important because its stores are greater and
because the C02 gain is constant over a wide range (Cherniack & Longobardo, 1973).
In support of such a view, these authors have pointed out that respiratory instability
follows a reduction in C02 stores, as following forced hyperventilation (Haldane &
Priestly, 1905), at altitude (Fitzgerald, 1913) when, in addition, the effect of hypoxia
starts to dominate, or the experimental reduction in gain of the C02 controller in the
adult (Cherniack, Euler, Homma & Kao, 1979) or new-born (Wennergen & Wennergen,
1980). Further, in both adult and new-born, respiratory instability may be reversed
by giving, as appropriate, C02 or High 02 to inhale.
Despite these uncertainties, we think that two points of some importance emerge
from the present study. First, it would seem that the form of analysis of respiratory
data developed here may prove to be a valuable method of exploring quantitatively
the factors which contribute to respiratory stability, e.g. by selective, reversible
interruption of known pathways in both adult and new-born animals. Secondly, the
present results may shed some light on the origin of some respiratory difficulties
encountered by a proportion of new-born infants in the days or weeks following birth.
Our results suggest that there is a period, after birth, which is longer than previously
thought, during which respiration is potentially unstable and, since only normal,
full-term infants were studied, this must be considered as physiological. We have
suggested that this instability is concealed over the first week or so, by the unusual
nature of the chemical disturbance: it becomes more obvious from the second week
onwards and it is only at a rather variable time after three months onwards that the
adult type of stability, with rapid respiratory recovery from a disturbance, is evident.
It remains entirely a matter of conjecture at present as to which of the factors,
outlined above, is responsible for these changes. The sequence is, however, consistent
with the observations of Fenner, Schalk, Hoenicke, Wedenburg & Roehling (1973)
and Hoppenbrouwers et at. (1977) that periodic respiration is not uncommon in a
proportion of normal, full-term infants for some weeks after birth but rare in the first
24 h.
Our results have also shown considerable variation in the degree to which cos 0 is
reduced after birth and also the time course of this reduction. It may be supposed
that, due to biological variation, there are some infants in whom development of full
chemical control of breathing after birth is even less complete or is further delayed
and in whom, in consequence, respiratory instability, as shown by episodes of periodic
breathing, would be yet more marked than in the present series. Such episodes need

15
VENTILATOR Y RESPONSES TO SIGHS IN INFANTS
not necessarily be regarded as dangerous since the majority will be self-limiting. They
may well be overlooked by parents or paediatricians and they will be most likely to
occur during non-r.e.m. sleep when the extrinsic and intrinsic respiratory drives
associated with, respectively, the awake state and r.e.m. sleep are reduced or absent.
This possibility would assume greater significance in those infants who have not
developed the mature type of response at 3-4 months of age since, at this time, the
time spent in non-r.e.m. sleep markedly increases (Parmalee, Wenner, Akiyama,
Shultz & Stern, 1967). In this context, it may be of significance that it is at this age
that the greatest evidence of sudden death during sleep, in otherwise apparently
normal infants, occurs (Naeye, 1980). The possibility that such disasters might be
prevented by identifying infants who might be vulnerable, using the techniques
described in this paper, could provide an additional motive for an extended study
of respiratory stability in the new-born.
These studies were generously supported by grants from Action Research - The National Fund
for Research into Crippling Diseases (A/8/1114), and the National Institutes of Health (PHS HD
09457). We would like to acknowledge the help of Dr Sue Evans of Bristol University Computer
centre, Dr Andrew Smith, Department of Physiology, Bristol University, and Mrs Zona Wilson.
REFERENCES

ANDERS, T., EMDE, R. & PARMALEE, A. (1971). A Manual of Standardized Terminology, Technique8
and Criteria for Scoring of State8 of Sleep and Wakefulnes9 in Newborn Infant&. Los Angeles:
U.C.L.A. Brain Information Service. NINDS Neurological Informative network.
ASKANAZI, J., SILVERBERG, P. A., FOSTER, R. J., HYMAN, A. I., MILic-EMILI, J. & KINNEY, J. M.
(1980). Effects of respiratory apparatus on breathing pattern. J. appl. Phyeiol. Reep. Environ.
Exercise Phy8. 48, 577-580.
BAND, D. M., CAMERON, I. R. & SEMPLE, S. J. C. (1969). Oscillations in arterial pH with breathing
in the cat. J. apple. Phy8iol. 26, 261-267.
BAYLISS, L. E.)'1966). Living Control Syetem8. London: English University Press.
BENDIXEN, H. H., SMITH, G. M. & MEAD, J. (1964). Pattern of ventilation in young adults. J. appl.
Physiol. 19, 195-198.
BIscoE, T. J. & PURVES, M. J. (1967). Carotid body chemoreceptor activity in the new-born lamb.
J. Physiol. 190, 443-454.
BRADY, J. P. & CERUTI, E. (1966). Chemoreceptor reflexes in the new-born infant: effects ofvarying
degrees of hypoxia on the heart rate and ventilation in a warm environment. J. Physiol. 184,
631-645.
BRYAN, A. C. & BRYAN, M. H. (1978). Control of respiration in the newborn. Clin. Perinatol. 5,
269-281.
BURGHES, D. & GRAHAM, A. (1980). Introduction to Control Theory including optimal control.
Chichester: Ellis Howard.
CHERNIACK, N. S. & LONGOBARDO, G. S. (1973). Cheyne-Stokes breathing. New Engl. J. Med. 288,
952-957.
CHERNIACK, N. S., EULER, C. VON., HOMMA, I. & KAO, F. F. (1979). Experimentally induced
Cheyne-Stokes breathing. Resp. Physiol. 37, 185-200.
CROss, K. W. (1954). Respiratory control in the neonatal period. Cold Spring Harb. Symp. quaint.
Biol. 19, 126-132.
DAVI, M., SANKARAN, D., MCCALLUM M., CATES, D. & RIGATTO, H. (1979). Effect of sleep state on
chest distortion and on the ventilatory response to CO2 in neonates. Pediat. Res. 13, 982-986.
DEJouRs, P. (1963). Control of respiration by arterial chemoreceptors. Ann. N. Y. Acad. Sci. 109,
682-695.
DRORBAUGH, J. E. & FENN, W. 0. (1955). A barometric method for measuring ventilation in
newborn infants. Pediatrics, Springfield 16, 81-87.

16

P. J. FLEMING AND OTHERS

EPSTEIN, R. A., EPSTEIN, M. A. F., HADDAD, G. C. & MELLINS, R. B. (1980). Practical implementation of the barometric method for measurement of tidal volume. J. apple. Phyaiol. Reap. Environ.
Exercise Phy8. 49, 1107-1115.
FENNER, A., SCHALK, U., HOENICKE, H., WEDENBURG, A. & RoEHLING, T. (1973). Periodic breathing in premature and neonatal babies: incidence, breathing patterns, respiratory gas tensions,
response to changes in the composition of ambient air. Pediat. Rem. 7, 174-183.
FITZGERALD, M. P. (1913). The changes in breathing and the blood at various high altitudes. Phil.
Trans. R. Soc. B 203, 351-371.
FLEMING, P. J. & LEVINE, M. R. (1982). Observations on respiratory stability in the new-born
infant. J. Physiol. 326, 58-59P.
FLEMING, P. J., LEVINE, M. R. & GONCALVES, A. (1982). Changes in respiratory pattern resulting
from the use of a facemask to record respiration in newborn infants. Pediat. Res. 16, 1031-1034.
FRANTZ, I. D., ADLER, S. M., THACH, B. T. & TAEUSCH, H. W. (1976). Maturational effects on
respiratory responses to carbon dioxide in premature infants. J. apple. Physiol. 41, 41-45.
GUTHRIE, R. D., STANDAERT, T., HODSON, W. A. & WOODRUM, D. (1980). Sleep and maturation of
eucapnic ventilation and C02 sensitivity in the premature primate. J. apple. Physiol. Resp.
Environ. Exercise Phys. 48, 347-354.
HALDANE, J. S. & PRIESTLEY, J. G. (1905). The regulation of lung ventilation. J. Physiol. 32,
225-266.
HATHORN, M. K. S. (1978). Analysis of periodic changes in ventilation in new-born babies.
J. Physiol. 285, 85-99.
HOPPENBROUWERS, T., HODGEMAN, J. E., HARPER, R. M., HOFFMAN, E., STERMAN, M. B. &
McGINTY, D. J. (1977). Polygraphic studies of normal infants during the first 6 months of life.
III. Incidence of apnoea and periodic breathing. Pediatrics, Springfield 60, 418-425.
KOWALIK, J. & OSBORNE, M. R. (1968). Methods for Unconstrained Optimization Problems. New
York: Elsevier.
KRAUSS, A. N., KLAIN, D. B., WALDMAN, S. & AULD, P. A. (1975). Ventilatory response to carbon
dioxide in newborn infants. Pediat. Rem. 9, 46-50.
LEWIS, G., PONTE, J. & PURVES, M. J. (1980). Fluctuations of Paco. with the same period as
respiration in the cat. J. Physiol. 298, 1-1 1.
NAEYE, R. L. (1980). Sudden infant death. Scient. Am. 242, 52-58.
OLSSON, T. & VICTORIN, L. (1970). Transthoracic impedance with special reference to newborn
infants and the ratio of air-to-fluid in the lungs. Acta paediat., Stockh. suppl. 207, 37-56.
PARMALEE, A. H., WENNER, W. H., AKIYAMA, Y., SHULTZ, M. & STERN, E. (1967). Sleep states in
premature infants. Develop. Med. & Child Neurol. 9, 70-77.
PHILLIPSON, E. A., MURPHY, E. A. & KozAR, L. F. (1976). Regulation of respiration in sleeping
dogs. J. apple. Physiol. 40, 688-693.
PURVES, M. J. (1966a). The effect of a single breath of oxygen on respiration in the newborn lamb.
Resp. Physiol. 1, 297-307.
PURVES, M. J. (1966b). The respiratory response of the new-born lamb to inhaled C02 with and
without accompanying hypoxia. J. Physiol. 185, 78-94.
RIGATTO, H., BRADY, J. P. & TORRE VERDUZCO, R. DELA, (1975). Chemoreceptorreflexesinpre-term
infants: the effect of gestational and post-natal age on the ventilatory response to inhaled CO2.
Pediatrics, Springfield 55, 614-621.
SCHWEILLER, G. H. (1968). Respiratory regulation during post-natal development in cats, rabbits
and some of its morphological substrate. Acta physiol. scand. suppl. 304, 7-123.
THACH, B. T. & TAEUSCH, H. W. (1976). Sighing in newborn human infants: role of inflation
augmenting reflex. J. appl. Physiol. 41, 502-507.
WAGGENER, T. B., FRANTZ, I. D., STARK, ANN & KRONAUER, R. E. (1982). Oscillatory breathing
patterns leading to apnoeic spells in infants. J. appl. Physiol. Resp. Environ. Exercise Phys. 52,
1288-1295.
WENNERGREN, G. & WENNERGREN, MARGARETA (1980). Respiratory effects elicited in newborn
animals via the central chemoreceptors. Acta physiol. scand. 108, 309-311.