Effect Measure

A curious paper on the 1918 flu pandemic appeared this month in CDC’s journal, Emerging Infectious Diseases. It seemed provocative, at least on the surface. It claimed that the conventional wisdom underlying pandemic flu preparations was wrong. It’s not the flu virus we should be defending ourselves against but the common bugs of the upper respiratory tract that take advantage of new fertile ground to grow in after the flu virus invades:

Medical and scientific experts now agree that bacteria, not influenza viruses, were the greatest cause of death during the 1918 flu pandemic.

Government efforts to gird for the next influenza pandemic – bird flu or otherwise – ought to take notice and stock up on antibiotics, says John Brundage, a medical microbiologist at the Armed Forces Health Surveillance Center in Silver Spring, Maryland.

Brundage’s team culled first-hand accounts, medical records and infection patterns from 1918 and 1919. Although a nasty strain of flu virus swept around the world, bacterial pneumonia that came on the heels of mostly mild cases of flu killed the majority of the 20 to 100 million victims of the so-called Spanish flu, they conclude.

“We agree completely that bacterial pneumonia played a major role in the mortality of the 1918 pandemic,” says Anthony Fauci, director of National Institute for Allergy and Infectious Disease in Bethesda, Maryland, and author of another journal article out next month that comes to a similar conclusion. (New Scientist)

It’s true that most flu experts agree that secondary bacterial pneumonias played a major role in the 1918 pandemic, so that’s not new. So what is new here? After examining the paper (you can find it here), I conclude: not much. And while it makes a point worth making, I’m not sure it makes it in the best way or draws the most important conclusion.

Here’s the main point of the argument. In 1918 an influenza virus swept the globe wreaking havoc and extraordinary morbidity and mortality. The case fatality ratio (CFR) has been estimated to be about 2%, give or take, which means that for every person with flu who died there were fifty who recovered. Since there were an awful lot of people who got flu in a relatively short period of time, an awful lot of people died in a short period of time as well. The fact that the number of cases and the number of deaths completely overwhelmed existing resources isn’t at all surprising. Currently our emergency rooms are taxed by ordinary flu seasons. A pandemic like 1918 would result in a catastrophic demand on services today, too, whatever the CFR. Brundage and Shanks, the authors of the EID article have a further point to make, however. There is a widespread notion that the 1918 virus was unusual, “hypervirulent,” that is, it wasn’t an ordinary flu virus but one that struck rapidly and lethally, producing a rapidly fatal primary viral pneumonia with accompanying acute respiratory distress syndrome (ARDS) accompanied by an acute dysregulation of the immune system in the form of a “cytokine storm.”

If you assume the main threat is a primary viral pneumonia then concentrating on means to prevent or treat the viral infection is the logical strategy, and indeed the medical response underlying pandemic plans emphasizes development of vaccines against possible pandemic influenza strains and antiviral agents like the neuraminisdase inhibitors, Tamiflu and Relenza. If the main problem, however, is secondary bacterial pneumonias then more emphasis would be given to vaccine against bacteria that cause pneumonia in flu patients (e.g., the pneumovax vaccine against Str. pneumoniae). This is a perfectly valid point and Brundage and Shanks go on to give reasons why they believe most deaths in 1918 were from secondary bacterial pneumonias rather than primary viral infections from a hypervirulent virus.

Their arguments are all perfectly plausible but highly circumstantial and in some cases not very weighty. For example, they consider the most cogent item the opinion of contemporary observers that most deaths were due to bacterial pneumonia, although these observations were made in an era prior to the discovery of the influenza virus. That a lot of smart people identified bacterial infection and isolated common respiratory tract bacteria from autopsy cases is neither surprising nor is it very informative. The same goes for virtually all their other arguments which rely on enough handwaving to generate a breeze that could knock you over.

Most people in 1918 who got flu didn’t die of it and the ones that did probably died mostly from secondary bacterial pneumonias. But now we have to ask what this has to do with today’s pandemic planning assumptions. In order to draw policy conclusions one has to make some additional assumptions. One is that a modern pandemic caused by the leading candidate, influenza A/H5N1 (“bird flu”), would be the like 1918 in character and severity. What we know about H5N1 to date is that differs from 1918 flu in two very important respects. One is that so far it isn’t very contagious. That’s the good difference. The bad difference is that so far H5N1′s CFR is a horrific 60% plus and much of the reason seems to be an associated ARDS. Even if the CFR is overestimated by a factor of ten (unlikely from the evidence to date, but possible) it would be three times worse than 1918. If the pathogenic mechanisms didn’t change — that is, if H5N1 behaved more like the small minority of 1918 cases that could be described as hypervirulent — then the magnitude of the task in treating and managing them would depend on what happens to transmissibility. That’s something we don’t know but it could potentially become much more transmissible while staying highly virulent (the idea that virulence and transmissibility must trade-off is wrong; they may or may not). It’s the serious cases that will tax the health care system and the society.

Still, the point is valid: don’t make the preparedness strategy all about the virus. But here’s where we start to part company with the authors. They suggest adding antibiotics and antibacterial vaccines to the mix because secondary pneumonias will be important, putting in place appropriate surveillance of antibiotic resistance and species prevalence. But that won’t solve the main problem which will be coping with the medical surge and the effects of widespread absenteeism. It’s not all about proper medical therapy, either. That’s important but if the medical care system isn’t able to handle most cases, also irrelevant. The medical care system and many other vital services is so brittle it won’t take much of a pandemic for them to break.

So point taken. But not far enough.

Comments

#1 PalMD
August 6, 2008

Great post!

Given that the large number of dramatic deaths from the pandemic were likely from ARDS, preparation involves prevention and resource management, among other things. The type of argument given above seems short-sighted as you pointed out—expanding the penumbra of preparedness cannot start by taking focus away from our most effective tool.
#2 Nigel Thomas
August 6, 2008

Good article. We need to keep Bird Flu at the forefront of every business manager’s mind. It won’t go away so better start preparing.

Nigel Thomas
For free references and tools go to Bird Flu Manual Online or, if you need more comprehensive tutorials and templates, consider Bird Flu D-I-Y eManual for business preparedness and survival.
#3 speedwell
August 6, 2008

Revere, you (or some of you) and I have had differences about certain issues in the past, but when it comes to bird flu I am solidly behind you and consider you one of my most valuable sources of straight, evidence-based information. Thank you very much for your coverage.
#4 caia
August 6, 2008

Most people in 1918 who got flu didn’t die of it and the ones that did probably died mostly from secondary bacterial pneumonias.

You know, I’ve been following H5N1 for years now, and hearing about the 1918 pandemic, and I hadn’t known that was the conclusion the experts had drawn.

Learn something basic every day, I guess.
#5 SusanC
August 6, 2008

Thank you, revere, for your analysis of the issues. Yes, this piece of the puzzle is a not very satisfactory one, because you can say those who died in 1918 died ultimately from secondary bacterial infection, but the W-shaped mortality curve begs this question: why should perfectly healthy young people succumb disproportionately to bacterial pneumonia? The W-curve persisted till 1922 or thereabouts. Which IMHO is a good argument for this virus having some unusual features that may have weakened or otherwise altered the host-virus immune interactions, causing them to be much less able to defend themselves against bacterial infection. So what is this something, this x-factor you might say, that created such a pattern? And are we seeing the same thing in H5N1 now? Is this characteristic of primary infections from pure avian viruses, once they acquired the ability to enter host cells?

What I find most intriguing, is the Imai paper (in Cell) that you explored a while back, where they were able to induce changes related to acute lung injury, by using inactivated H5N1 viruses. If those results are correct and can be replicated, are we seeing a whole new mechanism of pathogenesis that we never even considered in viral ‘infections’? Could it be that the notion of ‘infection’ ie entry and replication has overshadowed our ability to see things differently?

If an inactivated H5N1, which is unable to replicate (can it enter cells? I don’t know), can cause similar lung damage than the actual virus, we may be looking at some early stages where new types of treatments such as immune modulators may be possible. I am of course speculating and definite understanding is a long way off, but we may do well to start considering H5N1 disease as multi-factorial, and developing combination therapies eg statins/fibrates + antivirals + antibiotics. Maybe. Just MHO, FWIW.
#6 SusanC
August 6, 2008

ps, yes, neither combination nor indeed any treatment will solve the surge capacity issue. That is or ought to be the single biggest driver for action for tptb. The gap is so big even in a moderate pandemic, that we will need everything we can throw at it, like homecare, NPI, drive-thru clinics, off-site facilities etc etc. No room for complacency, in any part of the equation, again JMHO.
#7 Ren
August 6, 2008

I agree with you that there is nothing new in the paper. Still, seasonal flu and it’s pandemic cousins are a threat that must be watched. Maybe not all resources should be devoted to it, but some. Good post. Cheers.
#8 Ewen Callaway
August 6, 2008

In my reporting of this story I discovered how un-new their conclusion was and tried not to hype that aspect too much. But two peer reviewed articles in one month were enough to make me bite (Tony Fauci has a JID paper on the same topic coming out soon). And many of my — and perhaps even your — readers didn’t know that bacterial pneumonia played a major role in 1918, so this was a good chance to get that idea out. And the policy implications for pandemic influenza pushed this otherwise not-so-newsy story over the edge.

Thanks for the link, and keep up the good work.
#9 pft
August 6, 2008

Every book I have read (4) on the 1918 pandemic has made this point. Thats why even a layman like myself had pointed out in earlier comments that antibiotics should be given out prophylactically to anyone coming down with influenza in a pandemic rather than waiting for an overtaxed HC system to make a diagnosis. The emphasis on vaccines for preventing secondary bacterial infections in a pandemic does not make sense to me. We have antibiotics, just make sure we have enough and can readily distribute them in a pandemic without unnecessary delay.
#10 M. Randolph Kruger
August 6, 2008

Revere, it would seem to me and a few others on the back channels that bacteria does account for some of them. But so far it would seem that they were dying of organ failure mostly due to the lack of oxygen and that is caused by initial inflammation by H5N1. It also is able to penetrate the organs now, soI dont understand how they can make the statements without any real proof..

You might have access to this as well but if someone is coughing up green funk then antibiotics are automatic right? But surely someone has tried this by now because green is mean? Did it change any outcomes?

If the amount of stuff sloughing off is going in faster than it can be cleared wouldnt the deaths occur anyways or is this another try everything approach? I havent seen anything so far that would indicate that antibiotics would work, only that 25% or so of the survivors die within 2 weeks after getting rid of the flu because of all the crud in the lungs. (Tropical Disease Hospital-Wellcome UK).

Anything to add?
#11 revere
August 6, 2008

pft: the authors of this study recommend pneumovax vaccine for bacterial pneumonia and it makes sense to prevent a secondary infection than treating it with ntibiotics (to which it may become immune and for which there may be serious side effects). The vaccine strategy they were cautioning against was one against the virus, not against bacterial secondary infections.

Randy: Don’t confuse the subject of this paper, the 1918 pandemic, and what might or might not happen or what is happening now with H5N1. Most deaths in 1918 were from bacterial secondary infection. I think the evidence is pretty good on that. What is unclear is how a pandemic H5N1 or other pandemic flu virus will operate, however. At the moment the virus does seem to be hypervirulent (using their terms), although some questions about that remain as well as we don’t have the kind of clinical data we would like. The point of the post was that even if only a small fraction of the serioius pneumonias is of the ARDS variety it will break the health care system and that’s what we need to fix, not to mention the rest of normal functions in civil society upon which we dependl. You won’t fix them with stockpiles of antibiotics or pneumovax, although having adequate supplies of them is a necessary but not sufficient condition.
#12 Dylan
August 6, 2008

I’m really becoming increasingly weary of correcting the numbers that revolve around this subject (I’ve done this at least five times, to date, at different venues); especially when the numbers involved are almost universally acknowledged; and especially since the math involved is very basic, and very simple. But, oddly enough, I had to make the same correction when the head of HHS trotted out his chief scientific advisor (over on the government’s panflu blog), and he promptly stumbled over the numbers, just like everyone else.

This is how it is (and if anyone here has trouble understanding this, then all I can say is that you are in desperate need of some remedial math instruction):

In 1918 there were roughly 1.8 billion people occupying this planet. That is 1,800 million. Got that? If the CAR of the H1N1 virus amounted to 30% of the population (the figure most often cited), then the number of people infected was approximately 540 million. No? Yes. If you accept the lowest published number that I have ever seen, as the actual raw number of deaths that this pandemic produced (20 million), then you have a CFR of 3.7% (20 divided by 540; when each of these numbers are considered to be millions, then the outcome is 19.9 million; as I said, this is very simple math). Now extend the numbers to match the contemporary, accepted numbers. Every re-examination of the numbers tends to boost the number of fatalities. The current estimate runs from 40 million (considered a conservative estimate), to as much as 100 million. The latter figure is more likely to be an accurate assessment of the number of fatalities, than the former number. Why? Because any projection would require a more accurate assessment of the numbers encountered in those countries most capable of keeping reliable records during the course of the pandemic, and extrapolating from these figures to arrive at comparable figures for those countries that likely would have suffered a greater numbers of casualties, based simply on a comparison of extant population figures should produce a reasonably reliable figure.

Forty million dead equates to a CFR of 7.4%. From what we know today, I would consider that a very firm, bottom-line assessment of global mortality (assuming a 30% CAR, of course). And 100 million dead? That gives us a CFR of 18.5%. I think that this figure is realistic. And I think that it is entirely reasonable to assume that an H5N1 generated pandemic is completely capable of producing — at the very least — the same outcome. The biology appears to be largely parallel, if not entirely synonymous. We don’t know what preceded the pandemic, with regard to H1N1. We know, very intimately, what H5N1 is doing. I think that our current knowledge of H5N1 probably tells us what H1N1 did, in the beginning. That is not an unreasonable conclusion. I also think that our knowledge of H1N1 likely tells us what H5N1 will do, in the end. That, also is, in my opinion, at least, not an unreasonable conclusion.

Now, take the above figures for CAR and CFR, and factor them in to a population of 6500 million, rather than 1800 million. Assuming that a pandemic of H5N1 will only be as ferocious as the 1918 H1N1 pandemic. Very bad numbers. Far worse than anyone in a position of power is willing to discuss.

Revere: Please examine these numbers. I assure you that they are correct.
#13 Achak
August 6, 2008

…we may do well to start considering H5N1 disease as multi-factorial, and developing combination therapies eg statins/fibrates + antivirals + antibiotics. Maybe. Just MHO, FWIW.
Posted by: SusanC

Do we really have the choice to not heading that Way?
#14 Achak
August 6, 2008

In considering H5N1 disease as multi-factorial, and developing combination therapies eg statins/fibrates + antivirals + antibiotic aren’t we considering Bernard La Scola and colleagues from the Universite de la Mediterranee in Marseille who are reporting in the journal Nature that that Sputnik was able to achieve a remarkable degree of gene mixing by “looting” genes from its host virus and other organisms.

Viruses are already known to infect and sicken bacteria but this is the first example of a virus infecting one of its own kind.

Sputnik is the first example of a virus infecting another virus to make it sick.
#15 revere
August 6, 2008

Dylan: I’m not particularly concerned with the difference between 2% and 8% because either one is a catastrophe. The main point (for me, anyway) is not the total number (an exercise that is possible but I think we have insufficient data except to talk orders of magnitude) but the fact that even the lowest estimates point to a gargantuan problem — and indicate the only way to mitigate it, whether CFR is 2% or 8% o 18%. That is to make our communities more robust and resiient and to plan for and encourage the kind of behavior and resources that would make people pulling together to get through it easier. For me, the numbers are like nuclear overkill.

Achak: Link?
#16 Dylan
August 6, 2008

Revere: We have been, essentially, in agreement, in the course of the last four years. No dispute, about that. Where the “Public Health” issue is concerned. But 2% is simply not 18.5%. No matter who assesses it. And, judging by the nature of how the 1918 experience is concerned, 18.5% might very well be the “least” that we are looking at, here. Regardless of what the government might have to say about the matter. In excess of 60% CFR, worldwide, and above 80% in Indonesia. True?
#17 revere
August 6, 2008

Dylan: I’m not really disagreeing with you or your numbers. I am probably more hesitant about them than you are because all of them are based on some gross estimates, not just of the past, but the future. Yes, there is an order of magnitude difference between 2% and 18% that I would consider important so I grant you that point. While my gut tells me we aren’t looking at 20% CFR or even 10% CFR if there is an H5N1 pandemic, I wouldn’t want to plan on my gut instinct (why should anyone trust my guess rather than yours — and I mean that quite seriously), and your calculations are perfectly plausible on the basis of what evidence we have (but then so are a lot of other scenarios perfectly plausible). Still, I think that whether it is 2% or 18% I wouldn’t suggest we do anything different, which for me is the bigger point. And we aren’t doing anything to adequately prepare for either. Put it down to my pre-occupations.
#18 Achak
August 6, 2008

On Lascola

http://ifr48.timone.univ-mrs.fr/files/CV/Lascola-Bernard.pdf

His Contributions

http://en.scientificcommons.org/bernard_la_scola
#19 RobT
August 6, 2008

I don’t think SusanC’s question has been addressed properly. Revere, can you comment on the point she made about the W curve, please? That is, why would prime of life victims die disproportionately from bacterial infection, not the more elderly and the very young? Is this not counter-intuitive re bacterial pneumonia?
#20 anon
August 6, 2008

> I’m not particularly concerned with the
> difference between 2% and 8% because either
> one is a catastrophe

so, mitigating catastrophes makes no sense ?
#21 SusanC
August 8, 2008

RobT,

There is yet another piece of the puzzle for the W shaped curve, and that is, why should the mortality for older people be LOWER than their normal seasonal flu mortality? (here’s the chart from Taubenberger et al) One hypothesis put forward by Taubenberger and Palese (link) is that older people may have had partial immunity from previous exposure to a virus that circulated before 1989 when it was wiped out by a new virus. IF that is the case, it begs the question as to what the true CFR would have been for older people, if they were truly immunologically naive.

Here I put their 2 charts together for a better visual comparison. The idea put forward by Palese is very sobering. The very high CFR in various remote places in Alaska, Pacific islands and some parts of Africa (some 70-100%, Killingray et al) gives one pause, as to whether that would have been the pattern worldwide in a genuinely immunologically naive population.
#22 AnnieRN
August 8, 2008

And as a follow-up to RobT’s question: What about situations where the person sickened in the monring and was dead by nightfall? Is that still bacteria-related?
Great post, Revere. Thank you. AnnieRN
#23 revere
August 8, 2008

Couple of things. Regarding Susan’s question, the paper explains the W shape as being a consequence of much greater crowding in the young adult group (soldiers, for example) that led to increased transmission of bacterial pneumonia among people with weakened immune defenses because of the viral infection in the lung. The older age group had lower infection rates but higher mortality, presumably because us old folks have lots of other risk factors for dying. That’s their explanation, anyway. It’s a bit of handwaving.

Regarding rapid deaths (Annie’s question) I think the authors would say they were primary viral pneumonias, not bacterial ones but the data they present suggests this was uncommon, even for primary viral pneumonias whose interval from onset of early pneumonia to death was similar to those who got pneumonia late, i.e., getting a viral pneumonia didn’t seem to hasten death making the argument that the primary viral pneumonias were hypervirulent less likely. That’s their argument, anyway. Early deaths weren’t common they say and adduce evidence to show it.
#24 David L
August 10, 2008

Revere, could you perhaps review the public policy implications again. I, for one, am confused.

If secondary bacterial pneumonia was the cause of most of the 1918 deaths, wouldn’t the wide availability of antibiotics today mean that the risk of a new H5N1 pandemic be much less than previously postulated? For example, untreated typhoid fever has a CFR of 33% but it is rare for deaths to occur in the developed world because antibiotics are available and effective.

How much protection would the pneumomax vaccine offer? Based on that, should more people be vaccinated?

On another issue, how confident are we that the 1918 strain was a single variant? I would think that the early, allegedly highly lethal strains that swept through the Army hospitals and troop ships might quickly have mutated to a less lethal strain (killing your host is not a very successful reproductive strategy). Do we know how long the incubation periods were for the 1918 flu?

I’m no expert, but these seem like relevant questions.

David
#25 revere
August 10, 2008

David L: I think that one implication is that organizing preparedness around fighting the virus may not be the most reasonable alternative, given we are at this point unlikely to get a vaccine in a timely fashion, at least for wave 1. It is certainly a major part of the mix, it just isn’t the whole thing. The authors of the paper advise having a sufficient stockpile of antibiotics to treat secondary pneumonias as well (they point out that these drugs are part of the just in time inventory problem) and using bacterial vaccines like pneumovax (which I got along with my flu vaccination this year), although it won’t stop all secondary bacterial penuimonias, only those from S. pneumoniae.

My problem is that the main issue will be the medical surge which will overwhelm both the medical care facilities and the usual infrastructure that keeps “normal” modern life going. If we want to deal with that — and we do — then we need to pour substantial investments into the public health and social service infrastructure and prepare the other parts of civil society for the eventuality of high absenteeism, from whatever cause.

Regarding mutation to lower virulence, yes, it could happen. But it doesn’t have to happen. It didn’t happen with HIV or smallpox or Ebola and on theoretical grounds it isn’t mandatory (I have a link in the post to an earlier post on the science of it).
#26 Mary
August 10, 2008

Interesting discussion. I became interested in this subject as a result of experience with canine influenza, which has a lot of similarities; cause of death we saw was from hemmorhagic pneumonia which has a bacterial component that responded to antibiotics if they were started early enough.

As to overwhelming the medical system, if we believe antibiotics will be useful, why not cross-train other types of workers to respond to such an emergency and give them specific authority to act.

I’m not a medical professional, but I’ve learned to give animals vaccinations and sub-q injections. If I can learn that, anybody can. Other groups of people – lawyers, secretaries, farmers, plumbers – could be trained specifically to respond to one disease pattern and administer appropriate treatment in a pandemic situation. They would just need specific training and access to appropriate medications.
#27 revere
August 10, 2008

Mary: Lots of other stuff here (check out the categories on the left sidebar). For pandemic prepping take a look at the The Flu Wiki and Get Pandemic Ready as well as links you find there (NB also Dem’s post at DailyKos today).
#28 Jonathon Singleton
August 11, 2008

Revere, “…organizing preparedness around fighting the virus may not be the most reasonable alternative, given we are at this point unlikely to get a vaccine in a timely fashion, at least for wave 1…”

Revere, interesting news ’bout ProMED-mail deciding to close down reporting suspected human H5N1 infections/deaths in Indonesia until “official” confirmation by the Indonesian Ministry of Health. Which leads into your above comment on a vaccine based on wave 1 pandemic virus…

What is your opinion on prepandemic H5N1 vaccination? Quite obviously, there are a few trial-tested products around (eg. Australia’s CSL Panvax — see below). To me, the only aparrent issue is one of political will in both governments and WHO implementing the mass manufacture, distribution and public vaccination process…

Excerpt from an e-list sent email addressed to: “David W. Moskowitz MD, CEO, GenoMed” (Saturday, July 26, 2008)

Howdy Dave, I’ve just been reading through CSL’s prepandemic H5N1 vaccine, Panvax, trials conducted at Murdoch Childrens Research Institute, and the Melbourne School of Population Health, University of Melbourne, Carlton, Victoria, Australia (see PubMed).

I believe the results to be incredibly positive, constructed with an older Clade I strain but offering modest levels of cross-protection against the newer Clade 2 strains — the strains which surround Oz (see Table 2. Clade descriptions with isolation period, source, and geographic location from the CDC’s online report linked below).

So there is transgenic pandemic protection available for Australians — exactly how modest is that Clade 2 protection!?! All it would take is for the federal Oz government to [either dialogue with or] ignore the WHO and begin manufacturing and distribution.

Cheers:*) Jonathon

PubMed — Phase I and II randomised trials of the safety and immunogenicity of a prototype adjuvanted inactivated split-virus influenza A (H5N1) vaccine in healthy adults. 1: Vaccine. 2008 Aug 5;26(33):4160-7. Epub 2008 Jun 13

Excerpt: INTERPRETATION: Two doses of prototype 30mug or 45mug aluminium-adjuvanted, clade 1 H5N1 vaccines were immunogenic and well-tolerated with considerable 6-month antibody persistence. The prototype H5N1 vaccine also elicited modest levels of cross-protective MN antibodies against variant clade 2 H5N1 strains..

******************
Centers for Disease Control and Prevention — Toward a Unified Nomenclature System for Highly Pathogenic Avian Influenza Virus (H5N1) | CDC EID Volume 14, Number 7�July 2008

http://www.cdc.gov/eid/content/14/7/e1.htm
#29 bigTom
June 8, 2009

If surge capacity is going to be an issue, which is almost certainly the case, then mitigating the number of bacterial secondaries by vaccinating against common secondary bacteria should reduce the number of severe cases which will need the highest level of treatment. Any reduction in surge demand ought to be a good thing. I would think a figure of merit is the ratio of surge capacity to surge demand. That can be improved either by increasing capacity or by decreasing the likely demand. The optimal strategy likely demands a bit of both.