“Pulmonary Dead Space-ARDS” by Michael Matthay for OPENPediatrics

“Pulmonary Dead Space-ARDS” by Michael Matthay for OPENPediatrics


Good afternoon. My topic is an area that I’ve been interested
in now for 15 years, which is the potential clinical and physiologic significance of elevated
pulmonary dead space in patients with ARDS. I’m going to go back and give you the history
of how we became interested in this. First are my disclosures, and so then I’ll
go on from there. So pulmonary dead space in ARDS– five themes. First, background and rationale. Second, primary results from our initial study,
plus follow-up studies from us and other investigators. Mechanisms that might explain the elevated
dead space in patients with ARDS. And then the question of whether or not we
can or should estimate pulmonary dead space versus measure it. And then the conclusions. So everyone knows that the classic physiologic
abnormality in ARDS is hypoxemia. The patient’s chest x-ray, as shown here,
has bilateral airspace opacities. And the histologic picture shows alveolar
filling. And the hypoxemia’s secondary to alveolar
filling, and also some degree of atelectasis of the alveoli, resulting in low V/Q and frank
interpulmonary shunting. But we’ve noticed in ARDS network, and other
clinical trials, for many years that the baseline minute ventilation in patients enrolled in
clinical trials is virtually always elevated to about twice normal. So if normal minute ventilation’s about 6
liters, it’s consistently 12 liters. And this is in early ARDS. And the question is, why? We reasoned that either there is a marked
increase in carbon dioxide production or an increase in alveolar dead space. Now, many years ago, Julius Comroe from the
University of Pennsylvania, who founded our Cardiovascular Research Institute at UCSF,
published this simple diagram to illustrate how dead space really occurred simplistically
in the lung. And you can see here, on the side where the
B is, that we’re talking about is ventilation to an alveolar unit that does not have blood
flow. Now, there could be variations on a theme
here. There could be reduced blood flow as opposed
to no blood flow. But that’s the basic idea. And I always say, on rounds in ICU, what this
means is high V/Q. People are so used to saying V/Q mismatch. I think that’s an inadequate statement. I think patients should be described as having
a high V/Q, such as this, which will cause dead space, or low V/Q, which is more associated
with hypoxemia. So the hypotheses for our initial study are
shown here. The dead space fraction is elevated early
in ARDS. And, two, an elevated dead space fraction
will have an independent predictive value for identifying ARDS patients with a high
mortality. So to test this hypothesis, we did a study
over three years at both our University Medical Center and our city/county hospital in San
Francisco. 179 patients were prospectively enrolled,
and we collected clinical pulmonary and severity-of-illness data. At that time, we measured the dead space with
a bedside metabolic monitor, the DELTRAC– normally used to help guide nutritional replacement. We of course calculated the dead space fraction
with a modified Bohr equation, where in the numerator, you have arterial PCO2 concentration
divided by mixed expired CO2– not to be confused with end tidal CO2, but mixed expired CO2–
divided by the arterial PCO2. So in us, right now, my PCO2 is, let’s say,
40. My mixed expired CO2 is 30. If you put 40 in the denominator, it’s 10
divided by 40, or 25%, as a proximate normal dead space, reflecting largely anatomic dead
space. Tidal volume in this trial was a mean of 10
+/- 1.4 mL per kilogram. We were just at the cusp, the turnover period,
to low tidal volume, an issue I’ll come back to. And we also measured static respiratory compliance. Patients were a mean of 48 years old, a little
younger because of the inclusion of patients from the city hospital. Clinical disorders– sepsis, 25%; pneumonia,
31%; aspiration, 11%; major trauma, overdose, and so on, 34%. Overall mortality, 42%. Now here are the baseline characteristics. P/F was 147, pretty severe. Tidal volume, as I mentioned, 10 mL per kilogram,
ideal body weight. Respiratory compliance. Minute ventilation, just as we anticipated,
was twice normal, 12.1 liters. Dead space fraction overall was 0.58, so quite
elevated if a normal, as I’ll show you, is around 0.35 to 0.38 in a normal ventilated
patient without lung disease. Statistical analysis was done as the primary
outcome variable of death before hospital discharge; logistic regression; SAPS II for
severity of illness; and multiple logistic progression for independent association with
death. Now, at the end of the study, there were several
univariate variables that were associated with death, which you would expect. And let’s just go through them briefly. The P/F ratio was lower in non-survivors. The dead space was 0.63 in non-survivors and
0.54 in survivors. Respiratory compliance was lower in those
who died. Lung injury score was higher in those who
died. SAPS II was higher. And as you would expect, sepsis was higher,
51% versus 31% in those who died. Vasopressor use was about twice higher. The low tidal volume protocol was just making
its way into clinical practice. Just a small number of patients– 31– but
of course, it makes sense that it was more common in the survivors. Cirrhosis– not quite significant. And NPOFC is Non-Pulmonary Organ Failure. So then we put this into a multivariate analysis,
and the three variables that turned out to be independently predictive of mortality are
here. Pulmonary dead space fraction, with the highest
odds ratio– it had a 0.002 value; SAPS II, index of severity of illness; and respiratory
compliance. All the others fell out. And looking at the data just in a figure,
you can see here the early elevation of dead space. And these were done within 24 hours of the
diagnosis of ARDS. And you can see that the median was 0.58 in
ARDS versus ventilated controls, where it was 0.36. And here’s the data for non-survivors versus
survivors. It was significantly higher in the non-survivors
versus survivors, as I’ve already said. And you can see there’s a pretty good separation
there. And then when we broke the data down according
to dead space fractions, taking increments as shown in this figure, with mortality on
the y-axis and dead space fractions on the x-axis, you can see, at the cutoff, around
0.57 to 0.61, the relationship to mortality began to increase sharply in those two bar
graphs. And then in a follow-up study published two
years later by Rich Kallet one of the key authors of this work, we found that measuring
dead space at later time points in ventilated patients with ARDS still had predictive value
for death. Perhaps not a surprise, but it’s shown here
for day 1, day 2, day 3, and day 4. I think the most important observation was
day 1, but still, this shows there was a continuing association with a poor outcome. And then, in part because we wanted to see
if this still was true in the low tidal volume era, we did a follow-up study in 42 patients
with ARDS, this time measured by volume capnography, the simple bedside method that I’ll come back
to. All patients were ventilated with 6mL per
kilogram ideal body weight tidal volume. And the baseline pulmonary dead space again
was predictive of death– 0.61 versus 0.53. And several other studies have continued to
confirm this. Work from Dr. Pesanti’s and Dr. Gattinoni’s
group, published in the New England Journal in 2006, found that elevated dead space has
predictive value for mortality in ARDS. A study we did at UCSF, with Kathleen Liu
leading, showed that elevated dead space, when combined with a biologic marker, Ang-2,
was predictive mortality. And then there was this study in Chest in
2010 that also confirmed this. What accounts for this increase in dead space
fraction in early ARDS? Well, we don’t know precisely. But the two possibilities, it seems to me,
are thrombotic or inflammatory injury in the pulmonary capillaries, the microcirculation,
and/or obstruction of blood flow in the extra-alveolar lung circulation, resulting in areas of high
ratio of ventilation to perfusion. And many years ago, over 30 years ago, at
Mass General, there was a group led by Warren Zapol that was studying the pulmonary macrocirculation. And they did angiograms in patients in whom
they had pulmonary arterial catheters, and they wedged the catheter and injected dye. And they identified, in fact, macrovascular
obstruction in many of these patients. In fact, they found it in about half of their
patients, and it was associated with severe ARDS, presence of DIC, pulmonary hypertension,
and a poor outcome. Again, this is in the era before lung-protective
ventilation. And they also did very well-done studies of
patients who died and found obliteration of a lot of the lung capillaries in patients
who succumbed later, with thrombosis, medial thickening, and decreased vascular density. Now, the issue has come up– could we just
estimate dead space fraction, rather than measure it? And to answer this question, Jeremy Beitler,
of highly-motivated, bright pulmonary critical care fellow at Beth Israel in Boston, under
the supervision of Danny Talmor, worked with several of us, including Taylor Thompson,
to evaluate four different methods for estimating pulmonary dead space. The different methods are listed here, and
they were done in patients in whom we had data for, actual measurement of dead space
fraction. And I’ll just show you the results here. The manuscript is not yet published. I think it will be soon. But the main point to appreciate here is of
the four methods tested, only the Harris-Benedict one in the upper right-hand corner or left-hand
corner, actually had some reasonable relationship to the measurement of dead space. Harris and Benedict predicted measured VD/VT
to within +/- 0.10 in 70% of the patients. But that’s not really terrific, when you think
about dead space. If the dead space were 0.65, it could be wrong
in the direction of 0.55 or 0.75. So it’s not great. And if you took all 95% of the patients, the
standard deviation was 0.20, which is quite a lot. It also was less accurate when the PCO2 was
less than 30. So if you had to estimate, this would be the
one to use. But it’s not terrific. Further data in this slide just shows that
compared to all the others, the Harris-Benedict was the best for mirroring the actual measured
VD/VT association. It was within 7% of measured VD/VT in all
quintiles. It did improve predictive validity for mortality
when we added it to the Berlin definition. The Berlin definition didn’t have patients
in whom dead space was measured, and so they were hampered by that. There’s nothing they could do about it. But if we added this to the Berlin Definition
using this estimate, it did improve the mortality prediction. So it has some value, but our conclusions
are VD/VT should be measured in future randomized clinical trials to facilitate secondary analyses
to shape research and practice. The Harris-Benedict estimate best predicts
the measured VD/VT and association with mortality on a group level. It did enhance the predictive validity for
death over the Berlin definition alone, but I don’t think it’s adequate for individual
care and has the limitations I said. And just to close here, I would just would
like to make a plug for volume capnography. It’s very straightforward to measure. Right at the end of the endotracheal tube,
you get the mixed expired CO2, you get a blood gas, and you can measure the VD/VT straight
away. And this is data from Rich Kallet that shows
that with using one of the commercial methods– not that I’m plugging it, though– to measure
mixed expired CO2 with a NICO monitor. So in conclusion, elevated dead space fraction
predicts mortality and ARDS, a finding that has pathogenetic significance. Thrombotic and inflammatory injury to the
lung microcirculation is probably a major early mechanism of lung injury in ARDS. And I believe that measurement of pulmonary
dead space should be done in clinical research and trials in ARDS whenever possible. Thank you.

Leave a Reply

Your email address will not be published. Required fields are marked *