J Appl Physiol
91: 15821587, 2001.
PCO2 threshold for CNS oxygen toxicity in rats
in the low range of hyperbaric PO2
R. ARIELI, G. RASHKOVAN, Y. MOSKOVITZ, AND O. ERTRACHT
Israel Naval Medical Institute, IDF Medical Corps, Haifa 31080, Israel
Received 20 July 2000; accepted in final form 14 November 2000
Arieli, R., G. Rashkovan, Y. Moskovitz, and O. Er-
tracht. PCO2 threshold for CNS oxygen toxicity in rats in the
low range of hyperbaric PO2. J Appl Physiol 91: 15821587,
2001.Central nervous system (CNS) oxygen toxicity, as
manifested by the first electrical discharge (FED) in the
electroencephalogram, can occur as convulsions and loss of
consciousness. CO2 potentiates this risk by vasodilation and
pH reduction. We suggest that CO2 can produce CNS oxygen
toxicity at a PO2 that does not on its own ultimately cause
FED. We searched for the CO2 threshold that will result in
the appearance of FED at a PO2 between 507 and 253 kPa.
Rats were exposed to a PO2 and an inspired PCO2 in 1-kPa
steps to define the threshold for FED. The results confirmed
our assumption that each rat has its own PCO2 threshold, any
PCO2 above which will cause FED but below which no FED
will occur. As PO2 decreased from 507 to 456, 405, and 355
kPa, the percentage of rats that exhibited FED without the
addition of CO2 (F0) dropped from 91 to 62, to 8 and 0%,
respectively. The percentage of rats (F) having FED as a
function of PCO2 was sigmoid in shape and displaced toward
high PCO2 with the reduction in PO2. The following formula is
suggested to express risk as a function of PCO2 and PO2
F F0 100 F0/1 P50 /PCO2N
F0 0
P50 25.3 0.067 PO2 350 PO2 250
N e9.68 0.0252 PO2
F 0 234 0.637 PO2
P50 1.55 500 PO2 350
N 2.44
where P50 is the PCO2 for the half response and N is power. A
small increase in PCO2 at a PO2 that does not cause CNS
oxygen toxicity may shift an entire population into the risk
zone. Closed-circuit divers who are CO2 retainers or divers
who have elevated inspired CO2 are at increased risk of CNS
oxygen toxicity.
hyperbaric oxygen; electroencephalogram; convulsions; div-
ing; central nervous system
CENTRAL NERVOUS SYSTEM (CNS) oxygen toxicity can ap-
pear in humans exposed to oxygen pressures above 180
kPa as convulsions (similar to epileptic seizures, grand
mal) and loss of consciousness without any premoni-
tory symptoms. It is known that the risk of CNS oxygen
toxicity is greater when CO2 is present in the inspired
Address for reprint requests and other correspondence: R. Arieli,
Israel Naval Medical Institute, POB 8040, Haifa 31080, Israel
(Email: rarieli@netvision.net.il).
1582 8750-7587/01 $5.00 Copyright 2001 the American Physiological Society
gas (2, 7, 8, 13, 19, 21, 23) or in tissue (25), mainly due
to its effect on cerebral vasodilatation and increased
brain tissue PO2 (15) and acidity-enhanced reactive
oxygen species (ROS) (6, 11, 22). In underwater diving
where the risk of CNS oxygen toxicity is also encoun-
tered due to elevated PO2, other extraneous circum-
stances leading to the elevation of PCO2 may increase
this risk even more (20, 24).
We have recently shown (2) that the latency to CNS
oxygen toxicity decreased linearly as a function of the
inspired PCO2 and that it may even be affected by a
PCO2 as low as 1 kPa. These studies were conducted
using PO2 that cause CNS oxygen toxicity without the
presence of CO2 in the inspired gas. PO2 during closed-
circuit diving and hyperbaric oxygen therapy is usually
kept at levels that do not cause CNS oxygen toxicity.
All of the studies on the effect of CO2 on CNS oxygen
toxicity were conducted at a PO2 that can cause CNS
oxygen toxicity without the added effect of CO2 (7, 9,
10, 23). Almost nothing is known about the effect of
CO2 on CNS oxygen toxicity at a PO2 below the level
that will on its own produce this toxic effect. We hy-
pothesized that CO2, which causes vasodilatation in
the brain and therefore increases the tissue oxygen
pressure, together with acidity-enhanced production of
ROS (6, 11, 22), would cause CNS oxygen toxicity at a
PO2 below the level that causes CNS oxygen toxicity on
its own. The level of inspired CO2 at which the breach
occurs (the threshold for CNS oxygen toxicity) is the
subject of the present study.
The first electrical discharge (FED), which precedes
the clinical convulsions, in the electroencephalogram
(EEG) is a well-defined phenomenon and may be used
to validate the effect of PCO2 on CNS oxygen toxicity.
We used a previously described rat model (1, 3, 4) to
investigate the effect of CO2 in the inspired gas on the
threshold for CNS oxygen toxicity as a function of PO2.
METHODS
Animals
White male Sprague-Dawley rats had EEG electrodes im-
planted under equithensin anesthesia (0.3 ml/100g body wt
ip) 3 days before the experiment. The electrodes were stain-
The costs of publication of this article were defrayed in part by the
payment of page charges. The article must therefore be hereby
marked advertisement in accordance with 18 U.S.C. Section 1734
solely to indicate this fact.
http://www.jap.org
 PCO2 THRESHOLD FOR CNS OXYGEN TOXICITY
less steel screws that penetrated the skull in the parietal
area. Insulated wires attached to a female miniconnector
were soldered onto the screws, and the miniconnector was
fastened to the skull with dental cement. Ninety-one rats,
weighing 364 26 g (mean SD), were used in the study.
The experimental procedure was approved by the Animal
Care Committee of the Israel Ministry of Defense, and the
rats were handled in accordance with internationally ac-
cepted humane standards.
Experimental System and Procedures
Experimental cage. The experimental cage is a metal, dou-
ble-walled cage (25 20 12 cm; volume 6.0 liters). One
wall for observation of the animal and the top cover, which
can be opened, are made of Plexiglas. Thermoregulated water
can be pumped through the double wall to control the ambi-
ent temperature. The incoming gas flows through a metal
container attached to the cage wall for temperature equili-
bration before entering the cage itself. The metal walls of the
cage are covered on the outside with thermal insulating
material. A cable with a male miniconnector for EEG record-
ing passes through the top cover. Lin and Jamieson (18)
suggested that humidity in the inspired oxygen might affect
latency to CNS oxygen toxicity; therefore, the humidity in the
cage was monitored throughout the experiment. A humidity
and temperature measuring device (EE20FT, EE Electron-
ics, Linz, Austria) was inserted into the top cover of the cage.
Experimental system. The miniconnectors were mated, and
the rat was placed in the experimental cage, which was
placed in a 150-liter hyperbaric chamber (Roberto Galeazzi,
La Spezia, Italy). The flow of gas through the cage was
controlled by two needle valves: one controlling the flow of air
or oxygen and the other controlling the flow of CO2. The flow
was checked by observation of flowmeters situated inside the
hyperbaric chamber. The outgoing gas exited via a bypass
tube into the atmosphere of the hyperbaric chamber. A small
portion of the outgoing gas was directed out of the pressure
chamber (this was controlled by another needle valve),
passed through a flowmeter, and sampled by a CO2 analyzer
(CD-3A, Ametek, Pittsburgh, PA). Water hoses were con-
nected to ports in the pressure chamber and to ports in the
experimental cage for recirculation of the thermoregulated
water (C/H temperature controller bath and circulator 2067,
Forma Scientific, Marietta, OH). The temperature in the cage
was maintained at thermoneutral range to avoid any effect of
temperature on the CNS oxygen toxicity (1). EEG was re-
corded on a chart recorder (Gould, Cleveland, OH).
Experimental procedure. When the pressure in the cham-
ber was being raised (at 100 kPa/min), the gas flowing
through the cage was air. When the desired pressure was
reached, a period of 20 min was allowed for acclimation to the
experimental conditions, during which time air flowed
through the cage at 8 l/min and CO2 was added to produce
the desired concentration of CO2 in the cage. The flow of CO2
was adjusted by means of a fine needle valve to yield the
desired concentration, which was kept constant throughout
the exposure. After the end of the acclimation period, the flow
of air was immediately replaced by pure oxygen at a high flow
rate of 30 l/min for 1 min for fast replacement of the cages
atmosphere, while the flow of CO2 remained constant. The
flow rate was then reduced to 8 l/min, thus restoring the CO2
concentration within 1 min (a flow of 8 l/min through a 6-liter
cage). The EEG signal was amplified and recorded continu-
ously on a chart recorder. The fraction of CO2 in the inspired
gas, ambient temperature, and humidity were read and re-
corded. The rat was observed through a window in the pres-
J Appl Physiol VOL 91 OCTOBER 2001
1583
sure chamber. When FED, which precedes the clinical con-
vulsions, was seen on the recorder, decompression commenced
(at 100 kPa/min). The cage was removed from the pressure
chamber, and the rat was freed from the experimental sys-
tem. The time beginning 12 s from the start of high flow
oxygen (30 l/min) until the appearance of FED was noted as
the latency to CNS oxygen toxicity.
Experimental protocol. Each rat was subjected to a differ-
ent exposure (the combination of one PO2 and one PCO2) every
23 days. In previous studies (3, 4), we showed that the
preceding exposure has no effect on the time to FED in the
following one if there is a 2-day interval in between. If no
FED was recorded within 1 h of commencing the exposure,
that exposure was recorded as being without CNS oxygen
toxicity. Exposure to hyperbaric oxygen for longer than 1 h
may cause other forms of oxygen toxicity like pulmonary
oxygen toxicity, which, according to our experience, is ex-
pressed by heavy breathing beyond 1 h of hyperbaric oxygen,
and we assumed that most of the CNS oxygen toxicity events
would have a latent period of 1 h.
For each oxygen pressure, we determined the PCO2 range
within which lay the threshold for the CO2 effect. No FED
will occur at a PCO2 below this threshold, whereas the FED
will be produced at a PCO2 above the threshold. PCO2 at
intervals of 1 kPa were chosen for the desired resolution.
For the first few rats, this was done by trial and error in steps
of 1 kPa CO2. Later, PCO2 near the thresholds already found
for the first rats tested were chosen to minimize the number
of exposures required to determine the threshold for an
individual rat.
Six oxygen pressures were tested: 253, 304, 355, 405, 456,
and 507 kPa. At each pressure, various levels of inspired
PCO2 were selected in a procedure that searched for the PCO2
threshold. When a 1-kPa CO2 threshold resolution was suc-
cessful in any rat, this animal was tested again at another
PO2. The same animal was tested until the miniconnector
became detached, and, as a result, we were unable to test any
animal at more than three PO2 and usually at only one PO2.
PCO2 ranges that were selected during the experimental
procedure were 5.89.4, 1.19.7, 06.6, 04.4, 06.1, and
03.2 kPa for a PO2 of 253, 304, 355, 405, 456, and 507 kPa,
respectively.
Calculations. CALCULATION OF THE PERCENTAGE OF RATS EX-
HIBITING THE FED AS A FUNCTION OF PCO2. All measured data (rat
number, PCO2, latency to the FED in min, or 1 for no-FED)
for each of the five PO2, excluding those obtained at a PO2 of
507 kPa, were entered into a computer program. For each
PO2, the span of PCO2 over which FED occurred was deter-
mined from the lowest and highest PCO2 at which a rat
exhibited FED. Because PCO2 values were not grouped under
a few selected values but ranged over a continuum, we
divided the continuum into a few selected intervals to assess
whether the interval selection might influence the final re-
sults. Within this range of PCO2, intervals 5, 6, 7, 8, and 9
were selected one after the other. For each chosen number of
intervals, a table was constructed. Each row, representing
one rat, had either 1 for no-FED or the number of minutes
to FED at the appropriate PCO2 interval (columns). Because
the number of exposures per rat was usually less than the
number of intervals, there were empty cells at PCO2 intervals
for which there was no exposure. When two exposures of the
same rat happened to have their PCO2 in a single interval, the
last one recorded was entered into the appropriate cell. Thus
less exposures were used by the program as the number of
PCO2 intervals decreased because each interval spans a
greater range of PCO2. In a second step, the minutes were
converted to 1, and extrapolation and interpolation filled
www.jap.org
1584 PCO2 THRESHOLD FOR CNS OXYGEN TOXICITY

the empty cells. We assumed that the FED would occur at all
PCO2 above the threshold, whereas it would not appear at any
PCO2 below the threshold. Empty cells having 1 at a PCO2
lower then their own and no data above their own were
extrapolated to 1, and empty cells having only 1 at a PCO2
higher than their own and no data below their own were
extrapolated to 1. Empty cells having 1 on both of the
neighboring sides were interpolated to 1, and the same was
done with 1. The ratio of CNS oxygen toxicity (number of
1 signs) to the total (number of 1 and 1 signs) in each
column (PCO2 interval) and mean PCO2 were calculated. This
allowed us to present the percentage of rats having the FED
in the selected PCO2 intervals.
FED PROBABILITY EQUATION. Because many risk functions,
including the present PCO2 threshold for CNS oxygen toxic-
ity, are sigmoidal in form, we selected the versatile Hill
equation, which is used for sigmoidal responses in different
fields (e.g., hemoglobin oxygen saturation, dose response in
pharmacology, decompression risk), for the FED probability
equation. A sigmoid curve was fitted to the calculated per-
centage of FED as a function of PCO2 using nonlinear regres-
sion (SAS, Cary, NC). We assumed that at a relatively low
PO2 there should be a threshold PCO2 (D) and that when the
CO2 level exceeded this threshold it would cause CNS oxygen
toxicity
F F0 100 F0/1 P50 D/PCO2 DN PCO2 D on the overall response when the risks at different PO2
F0 PCO2 D
The parameters of percentage of rats that exhibited FED
without addition of CO2 (F0), PCO2 for the half response (P50),
threshold PCO2 (D), and power (N) were solved from the input
values of PCO2 and the percentage of rats (F). To minimize the
parameters for each statistical run, when the data showed
clearly that F0 was 0 (i.e., the FED occurred without any
CO2), D was deleted from the equation used for the nonlinear
regression. When the data showed clearly that there was a
threshold (i.e., no FED occurred at low levels of CO2), F0 was
eliminated from the equation. However, whenever the data
showed a clear D, the solved D was not different from 0 and,
therefore, was deleted. The apparent threshold is thus de-
scribed by very low values of F.
respectively. The difference in the latencies was signif-
icant (P 0.009, ANOVA). The distribution of mea-
sured latencies for the four lower PO2 is shown in Fig.
1 at intervals of 10-min bin. The distribution of the
number of measured latencies is largely affected by
PCO2 levels of 09.7 kPa as detailed in Experimental
protocol. However, at the four PO2, the number of
latencies is reduced from the 25-min bin to the 35-min
bin and to the 45-min bin and so forth. Even at the
lowest PO2, most of the CNS oxygen toxicity events
occurred within the first 40 min.
We did not use the PCO2 interval selection for 507
kPa O2 because virtually all of the animals (91%)
convulsed without any CO2, and, at any level of added
CO2, all of the rats had FED. The cumulative percent-
age of rats exhibiting FED for all six pressures, as a
function of inspired PCO2, is shown in Fig. 2 for both
intervals 5 and 9 PCO2. The lines connecting the sym-
bols deviate more from a smooth response with the
selection of nine intervals because each interval con-
tains less data. But the details of the sigmoid response
are better demonstrated by the nine-interval selection.
The selection of the number of intervals has little effect
are compared (Fig. 2). As the PO2 was lowered from 507
to 456 and 405 kPa, the percentage of rats exhibiting
FED without addition of CO2 was reduced from 91% to
62 and 8, respectively, until none convulsed at 355
kPa. At a PO2 lower than 355 kPa, the risk curve was

RESULTS

The results for each rat were tabulated in ascending

order of PCO2. Therefore, if a threshold is included in
the measured PCO2 range, no-FED may be expected
below the threshold, whereas a latency time will be
expected above this threshold. For example, when we
take rat 150 in PO2 of 304 kPa, there was no-FED at
PCO2 of 2.2, 3.8, 5.3, 6.6, and 7.1, whereas latencies of
21 and 38 min are seen at PCO2 of 7.5 and 9.7 kPa,
respectively. Thus the threshold for rat 150 at a PO2 of
304 kPa is at a PCO2 between 7.1 and 7.5 kPa. There
was a total of 381 hyperbaric exposures (51, 65, 70,
100, 63, and 32 exposures at 253, 304, 355, 405, 456,
and 507 kPa PO2, respectively) in 91 animals. The
mean number of exposures per rat was 4.7 3.5
(range 113). The combined mean humidity at the Fig. 1. Frequency of the latencies to the first electrical discharge
end of the experiment was 46.8 2.6%, and the ambi- (FED) preceding convulsions for 4 PO2. The latencies to the FED from
ent temperature was 27.5 0.4C. PO2 405 (A), 355 (B), 304 (C), and 253 (D) kPa were grouped into
10-min intervals. The number of latencies in each time interval is
The mean latencies to FED were 31.7 10.8, 26.0 presented as a function of the time interval. Only the data from the
7.2, 26.3 13.7, 20.9 9.2, 22.8 11.5, and 26.1 4 lower PO2 are shown to confirm that a 1-h exposure limit does not
10.9 min at 253, 304, 355, 405, 456, and 507 kPa PO2, exclude the longest latencies.

J Appl Physiol VOL 91 OCTOBER 2001 www.jap.org

 PCO2 THRESHOLD FOR CNS OXYGEN TOXICITY
Fig. 2. Percentage of rats that exhibited FED as a function of PCO2 at
6 oxygen pressures when either 5 (A) or 9 (B) intervals of PCO2 were
selected. The total number of rats that had FED at the highest PCO2
interval was taken as 100%. Then the percentage of rats with FED in
the other PCO2 intervals was calculated. For example, at a PO2 of 304
kPa (A), the rat that did not exhibit FED in PCO2 at the 3- to 4-kPa
interval and had FED in the 4- to 5-, 5- to 6-, 6- to 7-, and 7- to 8-kPa
intervals. Therefore, of the total of 25 rats (100%) in 304 kPa, the
4% related to this rat is added from the second interval onward.
shifted toward increasing levels of CO2. The overall
cumulative response seems to correspond to a sigmoi-
dal curve.
DISCUSSION
The hypothesis that elevated PCO2 will cause CNS
oxygen toxicity at oxygen pressures that do not on their
own induce toxicity has been proven correct. Our as-
sumption that there is a PCO2 threshold for each indi-
vidual rat, above which the rat will suffer CNS oxygen
toxicity and below which it will not, has been validated.
Of 381 measurements, there were only four cases in
which a rat had FED at a PCO2 below another PCO2 at
which it did not have FED. However, the difference in
PCO2 for these opposite responses was small (0.7 0.1
kPa), whereas for greater differences in PCO2 the re-
sponse of these rats was also in keeping with our
assumption.
The sudden switch from no effect to the development
of CNS oxygen toxicity may be ascribed to the effect of
CO2 on brain vasodilatation and elevated oxygen pres-
sure together with other effects that may be ascribed to
increased acidity. The extensive literature on ROS
postulates that acidity enhances ROS production.
Acidity affects the release of Fe2 that enhance hy-
droxyl radical production by the Fenton reaction (6,
J Appl Physiol VOL 91 OCTOBER 2001
1585
11). The hydrogen ion is a reactant in hydrogen perox-
ide production. The dismutation process is most rapid
at an acidic pH, and acidity enhances nitrogen oxide
and hydroxyl radical production (22).
In our previous studies on CNS oxygen toxicity in
rats, the exposure duration was limited to 1 h to avoid
other complications of oxygen toxicity (1, 3, 4). Those
studies were conducted at PO2 above 400 kPa. Because
the latency to FED increased as PO2 decreased (4), it is
possible that a condition that did not result in CNS
oxygen toxicity within 1 h will produce FED if the
exposure lasts beyond the 1-h limit. If, however, laten-
cies to FED longer than 1 h are expected, when we
search for the effect of CO2 near the threshold, laten-
cies close to 1 h may be expected there too. Figure 1
shows that, at all PO2 in the low range, most of the
latencies are below 40 min. The percentages of laten-
cies at the 40- to 50-min bin were only 8, 11, 3, and 12%
at a PO2 of 405, 355, 304, and 253 kPa, respectively; at
the 50- to 60-min bin, these percentages were only 6
and 8% at a PO2 of 355 and 253 kPa, respectively. We
therefore conclude that the 1-h limit did not affect the
results by excluding longer latencies.
Although the difference in the latencies to FED at
the different PO2 was significant, the values are close to
each other because of the additive effect of CO2 in the
low range of PO2. Thus the difference in latency be-
tween 507 and 253 kPa O2 was only 5.6 min compared
with a 19.3-min difference between 709 and 507 kPa O2
(4). We have previously shown that CO2 at various PO2
can reduce the latency to FED to the same minimal
level (2), which may explain the similarity here be-
tween the various PO2.
The cumulative probability of CNS oxygen toxicity is
most likely not related to random differences but to the
different inherent sensitivity of individual rats. We
have previously shown that the variability of sensitiv-
ity to CNS oxygen toxicity in the rat is greater between
than within animals (4, 5). Our data suggest that for
each individual rat there is a PCO2 threshold that
remains constant at least over the span of days during
which it was determined. To test for individual sensi-
tivity, we selected the rats that had been measured at
at least two oxygen pressures. We then scored all PCO2
thresholds at each oxygen pressure as its PCO2 interval.
PCO2 threshold scores were analyzed with repeated-
measures ANOVA. There was no significant difference
between the first and second oxygen pressures, but
there was a significant difference between rats (P
0.0009).
No probability equation was solved for FED at 507
kPa O2 because virtually all animals (91%) convulsed
without any CO2, and, at any level of added CO2, all
the rats exhibited FED. Thus F0 was the only param-
eter derived at 507 kPa O2. Nonlinear regression was
applied to the percentage of rats exhibiting FED as a
function of PCO2. The number of PCO2 intervals had no
significant effect on the solved parameters F0, N, and
P50. The solved FED probability equations for the six
PO2 are shown in Fig. 3. Superimposing Fig. 3 on either
panel of Fig. 2 presents a good agreement between data
www.jap.org
1586 PCO2 THRESHOLD FOR CNS OXYGEN TOXICITY

Fig. 3. Percentage of rats with central nervous system (CNS) oxygen

toxicity as a function of PCO2 (abscissa) and PO2 (values are written
on each curve). The lines represent the solution of the FED proba-
bility equation. The dotted curve for 507 kPa was calculated with
extrapolated parameters because only the intersection with the or-
dinate is known.

points and the lines. The curve for 507 kPa O2 (dotted)
was calculated from extrapolated parameters. As PO2
decreased from 507 kPa, the intersection of the sigmoid
curve with the ordinate moved down along the percent-
age of rats that convulsed. With a further reduction in
PO2, the whole sigmoid response shifted along the PCO2
axis to higher levels. At all PO2 except 304 kPa, most of
the response is completed within a range of 3 kPa CO2.
We do not have an explanation for the broader range at
304 kPa. The PCO2 above which F starts to rise can be
defined by a low value of F and not by a fixed threshold
(D in the method). If the value F 1% is chosen, then Fig. 4. The parameters of the FED probability equation (shown at
the thresholds for PO2 of 355, 304, and 253 kPa are at top) as a function of PO2. Equations that describe the dependent part
PCO2 of 0.3, 2.4, and 7.1 kPa, respectively. are also shown. N, power; F0, rats that exhibited FED without added
CO2; P50, PCO2 for the half response.
For a quantitative description of the risk of CNS
oxygen toxicity as a function of both PCO2 and PO2 (in
the low hyperoxic range), the parameters of the FED rameters (as yet unknown), this same model may be
probability equation were plotted against the oxygen used to help define the risk of CNS oxygen toxicity in
pressure (Fig. 4). P50 decreased linearly as the oxygen humans.
pressure increased from 250 to 350 kPa and remained When various agents that affect CNS oxygen toxicity
stable with further elevation of the PCO2. F0 was zero are compared at high and low oxygen pressures, their
when the oxygen pressure increased from 250 to 350 effect is more prominent at the low pressure. Thus
kPa and increased linearly as the oxygen pressure rose
above 350 kPa. N of the term (P50 /PCO2) (on a logarith-
mic scale) decreased linearly as the oxygen pressure
increased from 250 to 350 kPa and remained constant
with further elevation of PO2. These relations can be
summarized as follows
F F0 100 F0/1 P50 /PCO2N
F0 0
P50 25.3 0.067 PO2
N e9.68 0.0252 PO2
350 PO2 250

F 0 234 0.637 PO2

P50 1.55
N 2.44
500 PO2 350

Fig. 5. The slope of CNS oxygen toxicity risk (curves in Fig. 3) at

Because the physiology causing CNS oxygen toxicity is PCO2 P50, as calculated from the FED probability equation (shown
similar in humans and rats, with the appropriate pa- at top) and from the measured data adjacent to P50.

J Appl Physiol VOL 91 OCTOBER 2001 www.jap.org

 PCO2 THRESHOLD FOR CNS OXYGEN TOXICITY
increasing metabolic rate reduced latency to FED more
at 450 kPa than at 500 kPa (1), the effect of inspired
CO2 on shortening the latency is greater at 450 kPa
than at 500 kPa (2), and the effect of cinnarizine on
prolongation of the latency is more prominent at 500
kPa compared with at 600 kPa (5). To compare the
effect of oxygen pressure on the sensitivity of CNS
oxygen toxicity to CO2 at the CO2-affected range, we
calculated the slope (change in population with FED/
change in PCO2) of the sigmoid curve at P50 from the
derivative of the FED probability equation, and we also
measured the slope using the values adjacent to the
P50. The slope is shown in Fig. 5 as a function of PO2.
With the exception of the values at 304 kPa, the slope
increased as oxygen pressure decreased. That is, at low
hyperoxic PO2 in the PCO2-affected range, a smaller
increase in PCO2 will place more rats at risk of oxygen
toxicity.
Divers using closed-circuit breathing apparatuses, in
which the expired CO2 is absorbed within the system,
can be exposed to CO2 in the inspired gas due to failure
of the CO2 absorption unit. Professional divers have an
inherent or acquired tendency to hypoventilate during
strenuous underwater work and therefore have ele-
vated alveolar and tissue PCO2 (16). It is also possible
that divers with low sensitivity to CO2 are at an in-
creased risk of oxygen toxicity (17) as, for example, in
nitrox diving (14). When divers encounter increased
resistance to their breathing, the level of PCO2 in their
tissues will increase (24).
Closed-circuit or semiclosed-circuit diving appara-
tuses are designed for operation at oxygen pressures
below the level that can cause CNS oxygen toxicity.
Because all the studies of the effects of CO2 on CNS
oxygen toxicity have been conducted at oxygen pres-
sures that cause CNS oxygen toxicity without CO2 in
the inspired gas (7, 9, 10, 23), the present study com-
pletes the picture for low hyperbaric oxygen. This
study demonstrates that at a PO2 of 355 kPa, which
does not cause CNS oxygen toxicity in the resting rat,
just a small concentration of CO2 can turn a normal
scenario into a risk situation (Fig. 3). Such a possibility
should be considered in divers who are CO2 retainers
(14) or in conditions where high resistance may be
expected with elevation of CO2 (12, 20 , 24).
The authors thank R. Lincoln for skillful editing.
The research was supported in part by the Israel Ministry of
Health.
The opinions and assertions contained herein are the private ones
of the authors and are not to be construed as official or as reflecting
the views of the Israel Naval Medical Institute. The experiments
comply with current law in Israel.
REFERENCES
1. Arieli R. Latency of the oxygen toxicity of the central nervous
system in rats as a function of carbon dioxide production and
partial pressure of oxygen. Eur J Appl Physiol 78: 454459,
1998.
J Appl Physiol VOL 91 OCTOBER 2001
1587
2. Arieli R and Ertracht O. Latency to CNS oxygen toxicity in
rats as a function of PCO2 and PO2. Eur J Appl Physiol 80:
598603, 1999.
3. Arieli R and Gutterman A. Recovery time constant in central
nervous system O2 toxicity in the rat. Eur J Appl Physiol 75:
182187, 1997.
4. Arieli R and Hershko G. Prediction of central nervous system
oxygen toxicity in rats. J Appl Physiol 77: 19031906, 1994.
5. Arieli R, Shupak A, Shachal B, Shenedrey A, Ertracht O,
and Rashkovan G. Effect of the anti-motion-sickness medica-
tion cinnarizine on central nervous system oxygen toxicity. Un-
dersea Hyperb Med 26: 105109, 1999.
6. Baker MS and Gebicki JM. The effect of pH on the conversion
of superoxide to hydroxyl free radicals. Arch Biochem Biophys
234: 258264, 1984.
7. Bean JW and Zee D. Influence of anesthesia and CO2 on CNS
and pulmonary effects of O2 at high pressure. J Appl Physiol 21:
521526, 1966.
8. Chapin JL. Anticonvulsant threshold of CO2 in oxygen under
high pressure. Proc Soc Exp Biol Med 90: 663664, 1955.
9. Clark JM. Effects of acute and chronic hypercapnia on oxygen
tolerance in rats. J Appl Physiol 50: 10361044, 1981.
10. Clark JM. Interacting effects of hypoxia adaptation and acute
hypercapnia on oxygen tolerance in rats. J Appl Physiol 56:
11911198, 1984.
11. Crow JP, Spruell C, Chen J, Gunn C, Ischiropoulos H, Tsal
M, Smith CD, Radi R, Koppenol WH, and Beckmann JS.
The pH-dependent yield of hydroxyl radical products from per-
oxynitrite. Free Radic Biol Med 16: 331338, 1994.
12. Fagraeus L, Hesser CM, and Linnarsson D. Cardiorespira-
tory responses to graded exercise at increased ambient air pres-
sure. Acta Physiol Scand 91: 259274, 1974.
13. Hill L. The influence of carbon dioxide in the production of
oxygen poisoning. Q J Exp Physiol 23: 4950, 1933.
14. Kerem D, Daskalovic YI, Arieli R, and Shupak A. CO2
retention during hyperbaric exercise while breathing 40/60 ni-
trox. Undersea Hyperb Med 22: 339346, 1995.
15. Lambertsen CJ, Ewing JH, Kough RH, Gould R, and
Stroud MW III. Oxygen toxicity. Arterial and internal jugular
blood gas composition in man during inhalation of air, 100% O2
and 2% CO2 in O2 at 3.5 atmospheres ambient pressure. J Appl
Physiol 8: 255263, 1955.
16. Lanphier EH and Camporesi EM. Respiration and exertion.
In: The Physiology and Medicine of Diving (4th ed.), edited by
Bennett PB and Elliott DH. London: Saunders, 1993, p. 77120.
17. Leitch DR. An underrated safety aspect of CO2 in saturation
diving (Abstract). Undersea Biomed Res 6, Suppl: 2425, 1979.
18. Lin Y and Jamieson D. Effect of humidity on hyperoxic toxic-
ity. J Appl Physiol 75: 19801983, 1993.
19. Marshall JR and Lambertsen CJ. Interactions of increased
PO2 and PCO2 effects in producing convulsions and death in mice.
J Appl Physiol 16: 17, 1961.
20. Morrison JB, Florio JT, and Butt WS. Observations after loss
of consciousness under water. Undersea Biomed Res 5: 179187,
1978.
21. Shaw LA, Behnke AR, and Messer AC. The role of carbon
dioxide in producing the symptoms of oxygen poisoning. Am J
Physiol 108: 652661, 1934.
22. Shu Z, Jung M, Beger HG, Marzinzig M, Han F, Butzer U,
Bruckner UB, and Nussler AK. pH-dependent changes of
nitric oxide, peroxynitrite, and reactive oxygen species in hepa-
tocellular damage. Am J Physiol Gastrointest Liver Physiol 273:
G1118G1126, 1997.
23. Taylor HJ. The role of carbon dioxide in oxygen poisoning.
J Physiol (Lond) 109: 272280, 1949.
24. Warkander DE, Norfleet WT, Nagasawa GK, and Lund-
gren CEG. CO2 retention with minimal symptoms but severe
dysfunction during wet simulated dives to 6.8 atm abs. Undersea
Biomed Res 17: 515523, 1990.
25. Wood CD. Acetazolamide and CO2 in hyperbaric oxygen toxic-
ity. Undersea Biomed Res 9: 1520, 1982.
www.jap.org