Coronavirus (COVID-19) Press Conference: Marc Lipsitch, 04/30/20

You’re listening to a press conference from the Harvard T.H. Chan School of Public Health with Marc Lipsitch, professor of epidemiology and director of the Center for Communicable Disease Dynamics. This call was recorded at 11:30 am Eastern Time on Thursday, April 30.

Previous press conferences are linked at the bottom of this transcript.


MARC LIPSITCH: Welcome everyone. Looks like a smaller group today, which is perhaps a sign that there’s better communication coming from other sources which would be great because that’s one of the reasons we do these, but I’m going to just open it up immediately to questions. I don’t have any specific topics that I wanted to cover, so, I will take questions immediately.

MODERATOR: Great. Thank you, Dr. Lipsitch. First question.

Q: Thank you. Thanks very much. And thank you for doing this so consistent. I’m wondering to what extent you’re following the Massachusetts numbers and what we should make of the dynamics that we’re seeing. I have to go on the air very early tomorrow morning and give kind of a roundup of the week and I can – I’m just feeling very frustrated that the numbers are not going down or don’t seem to be going down more than they are. And I think I speak for the whole state by saying it’s hard to have the patience that we need to get through this. What would you say kind of speaking to just the Massachusetts audience?

MARC LIPSITCH: Yeah, thank you. I think I agree that it’s puzzling to see this plateau and Aaron Carroll wrote a nice article in the New York Times a week or so ago, pointing out that long plateaus are sort of hard to explain in models, which means that it’s hard to give them a mechanistic explanation account of why that should be happening. But we’re – although there’s been a lot of day to day variation. The sort of moving average seems to be kind of stuck at around 2000 cases or a little less per day.

It’s a little – I agree that that is a difficult situation to explain, other than to say that it seems that the reproduction number must be somewhere right around one and being held there and we haven’t gotten it down below one and it’s clear that some places can do that and how we’re going to do that I think is an ongoing puzzle. We’re not the only ones that are sort of stuck here, or stuck at some unacceptably high level in terms of level that we would be comfortable reducing our level of social distancing.

So, I have to admit to a bit of puzzlement myself and I hope that as the use of contact tracing begins to be possible that will help. I’m not actually that optimistic. I don’t think contact racing with 1000 or 2000 known cases per day is going to be very effective, because that means there are an awful lot of unknown cases and I just don’t see it as being by itself a big increment to our control measures. So, I think it’s – when next my colleague, Caroline Buckee is on one of these, I would ask her whether the Massachusetts mobility data is suggestive of sort of less effective social distancing then is seen in some other places. That would be my only real hypothesis for what it is and I haven’t talked to her in a couple days, so I don’t know whether that’s the case about Massachusetts. So, I wish I could give you a better answer, but I’m also puzzled.

Q: Well, comforting. At least I do have one quick follow up which is, is there a number that you watch the most? I feel like we’ve kind of been watching the hospitalization number of the most because we needed to make sure we didn’t overwhelm the capacity, but that’s staying pretty stable. Is there, is there any number – and yet the testing numbers are not very reliable because that depends a lot on how much you test, so what number do you watch the most?

MARC LIPSITCH: Yeah, I think it’s necessary to watch them all as you know hospitalizations is probably somewhat more reliable but it’s lagged because it takes some time for people to get sick enough to go to the hospital, to be admitted to the hospital. And so, I try to watch each of the cases, each of the indicators separately, but you really don’t know Hospitalizations gives you at least five or six days and maybe more than that, depending, in terms of lag depending on how fast the hospitalizations themselves get reported.

Q: Thank you.

MODERATOR: Okay, next question.

Q: Hi, Marc. I wondered if you might be able to talk a little bit about the work you’ve done on developing ways to accelerate vaccine trials ethically.

MARC LIPSITCH: Thank you. Yes. So, with several colleagues, we have written a proposal that’s published in the Journal of Infectious Diseases. Those colleagues are Nir Eyal, who used to be with us at the School of Public Health and is now at Rutgers, and Peter Smith at the London School of Hygiene. And our proposal is that in order to make more rapid evaluations of how well vaccines work and also evaluations that involve immunizing fewer people in the testing phase, we should consider the use of controlled human challenge trials in which individuals are given a vaccine or placebo, maybe different vaccines, there might be multiple arms and then are deliberately exposed to SARS CoV-2 infection. These would be volunteers who were screened, not only for truly informed consent, but also for having the lowest possible probability of developing complications if infected and preferably selected from areas where there is ongoing transmission anyway, so that their likelihood of being infected would be relatively high outside the study.

The rationale for this is that human challenge trials typically involve 10s to hundreds of individuals rather than the thousands who are involved in classical efficacy trials. I know from discussions with various individuals who are in groups that are trying to plan such efficacy trials that it’s quite uncertain where there will be adequate both infrastructure and incidence of infection at the time of the trial to do the trial well. And so, the controlled human challenge study has some logistical and timing benefits, potentially. It also in some sense is advantageous because it reduces the number of people who have to be exposed to the vaccine and the initial testing phase. And that’s beneficial because, as many of you are aware, there is concern with the with the vaccines against coronaviruses in general, including this one, about the possibility of so-called immune enhancement or disease enhancement by the vaccine, where the vaccine produces an immune response that actually is harmful to the individual once they’re exposed and while we obviously would not ever put a vaccine into trials where we thought that was very likely, we can’t be sure. And one of the benefits of a challenge model is that with 10s or hundreds of individuals exposed, if there is a signal of that you find it out earlier and actually jeopardize fewer people in that way through the trial. Again, that’s something we would try in every way to avoid but it is a further benefit of the approach.

Since we wrote that article, there have been several encouraging developments about the possibility of doing that there are discussions at the Wellcome Trust, at the World Health Organization, and 35 members of Congress wrote a letter to the FDA, which has prompted discussions within the US Government about this idea. There’s been a fair amount of press coverage and there’s also an effort that we sort of were supportive of that, but it’s an independent effort to evaluate how many people, or whether there are, in fact, people who might volunteer for such a trial. So, a group, an independent group, set up a website called and at the moment, there have been volunteers who have gone to that website and said, I would like to be in one of these trials.

And obviously bioethicists worry a lot about the concern that people would be sort of enroll in a trial like this because they’re desperate or because they need money, or because they are somehow in a vulnerable position and being exploited, and what’s remarkable is that many of the individuals who signed up on here leave comments and they are from all over the world, from 52 countries. They are often highly educated people who have a very strong altruistic motivation and wish to do this.

And one of the things that we write about in the paper is that we asked people, we invite people to take medical risks on the behalf of others in many settings. We do that when we ask someone to become a new physician who may be treating contagious diseases like this one, or Emergency Medical Technician or other type of dangerous profession. We also invite people to donate kidneys, which is a medical risk to them for someone else’s benefit, and obviously many kinds of first responders and military personnel are volunteering for the for that risk on behalf of others. And I think our rationale is in part that that’s a form of altruism that we rely on in society to keep our military and our healthcare system and other things going, and that there’s nothing completely different about a research setting where someone might wish to do that. But obviously, there are many ways that we can try to make it as low risk as possible as one would be obliged to do in other is really quite remarkable just as a sort of social phenomenon that this many people have volunteered.

Q: Thank you.

MODERATOR: Next question.

Q: Hi, thanks for doing the call. I’m just wanting to know what are you seeing in the southern hemisphere, as it approaches, it’s winter period.
Is a situation worsening and what might their experience show for us up in the US will serve as kind of a preview or offer some lessons for us as we prepare for a second possible wave?

MARC LIPSITCH: Well, the Southern Hemisphere is a big place, as you know, natural world and it’s quite variable. So, Australia and New Zealand seem to have achieved remarkable control of their epidemic, especially New Zealand. And I think what that’s shown is that although the spread started in the summer, what that shows us that clearly spread can occur in the summer, but also that it was caught relatively early and controlled fairly readily and I think, indeed, we will learn more. Well, so that’s Australia, New Zealand.

The phenomenon in Sub Saharan Africa is something that many people have been discussing the interpretation of and I think nobody is entirely certain why spread in Sub Saharan Africa appears to be relatively less severe than in many other places or than in many places in the Northern Hemisphere. Some of the hypotheses are that there were relatively few seeding events and so it’s just taking a longer time to reach high levels, and our research group has been doing some work on just how many seeding events there might have been, which we’re hoping to have out pretty soon. So that’s one hypothesis and other hypothesis is that something about the weather has been reducing transmission. And a third hypothesis is that just the health systems are not detecting the cases that actually exist. And I suspect there are aspects of all of those, but I think we’ve certainly learned that the Southern Hemisphere is not all behaving like the Northern Hemisphere and what the interpretation of that is I think remains debatable. And we really need better evidence.

Q: Just real quick, does that mean for New Zealand, Australia because of their efforts now, they might be in a much better position as the temperatures cool or do they have to work extra hard to avoid some kind of relapse if this virus behaves like other diseases?

MARC LIPSITCH: Yeah, I think it’s likely that as temperatures cool and humidity goes down, based on comparisons to other viruses and based on the fact that the other coronaviruses behave a lot like seasonal flu in terms of their timing, we would expect that they would have more transmission and by the same token, we can expect a little bit of reduction in transmission in the Northern Hemisphere.

My personal view is that the amount of trends – and I think fairly consensus view is that the amount of reduction in transmission from changing weather in the Northern Hemisphere is likely to be perhaps noticeable, although hard to disentangle from all the changing control measures, but not by itself enough to stop transmission in the Northern Hemisphere. I think it’s possible unlikely, but not completely impossible that the mixture of significant herd immunity in places like New York and the seasonal changes might mean that the next resurgence in New York is considerably milder in terms of rate of growth than it has been the first time because I think those two factors together might make a significant dent in transmission, but I think in places that have had more modest transmission, we might see the effects, but it might be very hard to be sure that they’re weather affects rather than composure effects.

MODERATOR: Thank you. Next question.

Q: Hi. Thank you, Dr. Lipsitch. I appreciate you taking the time to meet with us. My question is about contract tracing and really whether doing it halfway is better than not doing it at all. We’re trying to understand Florida’s contact tracing strategy and whether it’s sufficient and the health department tells us they’re hiring about 150 students to do it.

And I don’t know, is that level of effort better than nothing, because I saw that New York City and Boston are hiring a lot more people and casting a wider net, not just the students. So, the question I suppose is what’s the right way to do it and, if we have limited resources, where do we focus?

MARC LIPSITCH: Yeah, I think that’s a very good question and I was just on a on a grand rounds with Harvard Medical School this morning and there was a lot of discussion about sort of what is the level of fact we might get from contact tracing. I tend to be somewhat on the pessimistic side about contact tracing as a general control strategy. I think that it can be – well, I’ll start, I’ll end with my opinion and start with what we are more sure about. What we are pretty sure about is that contact tracing works well when you have relatively few cases so that you have the resources to do it, and when you have even fewer cases that are unknown. So, when you’re doing a good job of detecting most cases so testing capacity is high relative to their case burden.

Neither of those is true right now in most parts of the United States. As we just discussed with Massachusetts, we have a couple thousand cases we know about every day and, as the Chelsea study showed, probably quite a number of cases that we don’t know about. And I think that’s true in in many other places. So, I think the contact tracing in this current setting is going to be a fairly modest contributor, because it quickly can overwhelm especially 150 people in the state as big as Florida sounds a little bit like Austin Powers and the million dollars. And I probably should have said that on the record but I did. So, I think it’s easy to overwhelm a relatively constrained group of people trying to do contact tracing and it’s because it’s resource intensive. It diverts health, public health effort from other activities. On the other hand, I think, for certain settings like nursing homes or other settings where you really do need to try to identify exactly who’s infected and try to stop transmission as much as possible as fast as possible, you know, then, then it can be very useful in a targeted sense of trying to protect the vulnerable population.

It’s clear that you can do a lot of control if you do contact tracing really well. Singapore managed with mainly contact tracing for several months. But then eventually even Singapore lost its control of the epidemic and had to resort to social distancing types of measures. So, I think in conjunction with really aggressive measures to get case numbers down and significant resources, contact tracing can be a useful piece of the control approach, but I think it’s very challenging with an ongoing epidemic.

Q: Thank you, Dr. Lipsitch. I appreciate that.

MODERATOR: Next question.

Q: Hi, Dr. Lipsitch, thanks so much for doing this. So, we have – my question is about the remaining mystery of the virus origin. And clearly, there’s a lot of questions about where it really came from. And that conversation has of course also gotten politicized lately. And my question is what would a kind of healthy, transparent investigation look like? And is there a precedent for international cooperation around, you know, finding patient zero? And what could the international community do now, you know, to collaboratively figure out the origin of this virus in a hypothetical ideal way?

MARC LIPSITCH: That is a great question. I have been trying to stay out of that discussion, and I will describe the reasons why I’m staying out of it, but then I’m going to more or less stay out of it. As some of you may know, I have spent the last five years or so before COVID being quite active in the effort to discourage funding of and performance of gain of function experiments, the experiments to enhance the contagiousness or the viralness of a respiratory virus, mainly flu but it also involves coronaviruses to some degree. And I think that, at this point, we have a lot of immediate needs that the scientific community needs to do to address in a unified way to get to deal with the consequences of this epidemic.

The topic of origins and the policy implications for what sorts of experiments should be done are important topics and there are obviously very politicized and polarizing topics as you described. Within the scientific community, the debate was quite polarizing even before it became sort of part of poly party politics or geopolitics. For that reason, I am very anxious to know the answer, and I hope that processes will be undertaken and I’m going to try to stay out of this discussion and work together with colleagues, some of whom I strongly disagree with on those sorts of topics to deal with the problem before us, and perhaps we’ll get back to that later. But, I’m going to dodge the question completely because I really want to focus on the immediate problem before us.

Q: Okay, thank you.

MODERATOR: Next question.

Q: I wanted to just ask, in general, you know what you’ve seen from governors and states in terms of you know, starting to reopen economies, you know, in general, do you feel like governors are going too fast or about right in terms of the actions they’ve taken so far given the level of transmission still going on, the level of testing that we have that’s maybe not enough. I mean, what’s your kind of overall sense of how fast things are going?

MARC LIPSITCH: I actually am not sure that I have a really good overview of how things are moving. My sense is that some of the southern states are opening. I know George’s opening up some particular industries or types of establishments and some others, and in the southeast. But one of the consequences of this epidemic less up on the news than I would like to be.
My overall impression is that nowhere is there very good testing and that in the absence of much better testing capacity, it’s a dangerous thing to start lifting restrictions.

I’m perhaps a little less critical than some of my colleagues of those who are starting to reopen because I’m mindful that you know disease control is one exceedingly important piece of the puzzle, but it’s not the only piece of the puzzle. And while I tend to like the general policies of those who are being more cautious and therefore trust them more, I do have to say that, you know, unemployment is bad for people’s health, economic problems in general are bad for people’s health. And even if you only cared about health that wouldn’t mean you only care about COVID.

So, I’m somewhat sympathetic to the impulse to – as I think everyone is to the impulse to try to open up. I don’t think that at this point it’s good public health advice to reopen in most parts of the United States because case numbers are high and testing is poor. And that’s where we were a long time ago. And in particular, I think one thing that has been at least an idea I’ve heard express many times as well we’ve reached the peak and therefore it’s time to reopen but viruses don’t know where they were in the past, they only know where they are in the present. And what I mean by that is if we had x number of cases per day at the start of imposing restrictions and now we have a lot more than x cases per day, as seems to be the case in most places, but we’ve slowed down the increase, we’re in a worse position now than we were then. We have more virus spreading right now than we did at that time, assuming that it’s not all the testing artifact.

And so there’s not much logic in the idea that slowing down the growth is good enough and now we can reopen. The logic is in reopening when the number of cases has come down sufficiently that we can expect that the growth that will result when we reopen is manageable and will not overwhelm our healthcare system. So, if we were worried about it being overwhelmed a month ago, or a month and a half ago when the restrictions started and there are more cases per day now, then we should be more worried, not less.

Q: Is there a level of like an absolute level of cases in a state that you would look at and say, oh, if it goes below this many new cases a day, then that’s a good sign for reopening or something like that?

MARC LIPSITCH: Yeah, we’ve been trying to think through that issue. And some people have made proposals. I mean in qualitative terms, the idea is that you should be able to tolerate – you should have a level of cases such that if the level of cases went up by a large factor, say, five or 10, it would not overwhelm the healthcare system because it’s going to take – and that you have enough testing and monitoring in place so that you think you can estimate how much the level of cases is going up and calibrate the interventions to that rate of increase.

So, those aren’t hard numbers, but that’s in part because it varies by your hospital capacity and your testing capacity. So, to take to take two examples Massachusetts is getting much better on testing and has more testing that many places, but still has a large number per capita of new cases per day and a large number of hospitalizations. So, I don’t think we’re there yet. The country of Austria with 8 million people, so about the same size roughly as Massachusetts, is below 100 new cases a day, has relatively good testing capacity, and is thinking seriously about opening up. They stepped on the brakes earlier in their epidemic than most other places because they had a big outbreak in a ski resort and that brought attention. So, they are at a point where they could in fact deal with an upsurge in cases from their current levels and that’s a more sensible place to start thinking about reopening than here.

Q: Okay, thank you.

MODERATOR: Next question.

Q: Hi, thanks for taking my question. My question is actually related to the last one. You’re mentioning that you know you’re not seeing a lot of enough testing. So, I guess, what do you think is the best metric for evaluating that? Is that percent of positive tests or, you know, a lot of people are throwing around these national numbers that we need to have 5 million tests per day, even 20 million tests per day, others are saying, we just need to double, maybe triple what we have currently. So, what are your thoughts on you know those estimates and then is there a better way of thinking about it, rather than just like this national number?

MARC LIPSITCH: Yeah, I think this is another thing I would like to form a better opinion, a more better-informed opinion on and haven’t yet. But one thing to say for sure is that it’s very much a local issue in that if there’s a test available in another state, it doesn’t do me any good to a first approximation, and also you know, it’s very clear now from the serosurveys that very, very local conditions, even say within the Boston area, lead to very different levels of infection and disease. So, I think these decisions are very dependent on local conditions and I hope to have a more quantitative answer before too long but I don’t at the moment.

Q: Okay. And, but I guess, do you have a general opinion on just the very, very high numbers like 20 million tests per day? Is that completely out there, or do you think that is something that we would need to see or only later?

MARC LIPSITCH: I think it depends on how far we get the number of cases down. So yeah, I don’t know that I have a – I’m going to pass on that one because I don’t really have a good answer.

Q: Okay, thank you.

MODERATOR: Next question.

Q: Hey Marc. I’m wondering your thoughts on the antibody tests that New York is doing right now. The state put a bunch of resources and time into getting this test developed, kind of going to be the key to the reopening but now it looks like that’s not going to be the case, not going to be a key metric, especially with what the World Health Organization said about there being no proof that that’s going to prove immunity last week. Your thoughts on the effectiveness of that testing and how it should play into any reopening systems?

MARC LIPSITCH: Well, I think what the World Health Organization said correctly is that there is no evidence yet but everybody I know – I know
a particular set of the world of people in the world – almost everybody I know is working on trying to figure out how we’re going to get that information. So, I don’t think it’s a permanent state of ignorance, it’s a current state of ignorance, because this is a new disease and the studies to set that up to test that are actually quite challenging. And one of the things our lab is working on is the design and analytical approaches to doing those studies.

And just to divert briefly into that issue, you know, everybody says with epidemiological studies that are observational, so you don’t control who has the exposure and who doesn’t but you just watch and see who has the exposure and who doesn’t. You know, if that’s eggs per day and your outcome is heart disease, then you have to account for all the other potential causes of both eating eggs and getting heart disease. So, that’s the usual problem in epidemiologic studies and because it’s a natural infection that you get, we have that same problem, that we don’t decide who gets which exposure. So, it’s the classic problem in epidemiology.

What makes it particularly challenging in this case is that what we call the exposure, the thing that you’re trying to understand the effect of and the outcome, the effect you’re trying to understand, are the same thing. The exposure is infection now and the outcome is infection later, and so when you’re when you’re trying to list all the things that could cause both of those, there’s a long list because people who have who work in a supermarket are more likely to be exposed now are more likely to be exposed later than people who work at home. And similarly, people who have good personal protective equipment are less likely to be exposed now and later.

So, there’s all sorts of challenges in the design of these studies, and so, I think personally that that studies in healthcare workers, where you have reasonably sized cohorts of people who have similar levels of exposure because they have the same job for example, and either work with COVID patients or don’t knowingly work with COVID patients, those are going to be the best, I think, studies to study seroprotection, but they haven’t been – they’re just being set up now. So, I think we will begin to know and, frankly, I also think that someplace is going to do the experiment without waiting for the data, meaning do the experiment of sending people back to work with positive antibody tests and that probably through that, if protection is really good or really lousy, we may just find it out that way through an uncontrolled experiment. If it’s in between and subtle, it may be harder to figure that out. But if hundreds of people are sent back because of antibody positive tests to a place where there’s a lot of exposure and none of them get infected, that will be indicative and if on the other hand, lots of them get infected that will be indicative. So, you know, nobody could say that’s an advisable thing to do but the world being what it is, somebody’s going to do it. And it wouldn’t surprise me if the first data on seroprotection come from somebody just deciding to find out the hard way, using someone else’s exposure as the opportunity.

Q: Great, thanks.

MODERATOR: Next question.

Q: Hi, Dr. Thanks for taking questions. In the last week or two here in Washington, we’ve been hearing from Dr. Deborah Birx that what we really need on testing is going to be sort of rapid antigen test. And we’ve sort of gone through these cycles. First, everything was fixated on PCR capacity than antibody tests, which have all the issues you discussed.
What’s your view of the value of an antigen test? And do you have any thoughts on why those seem to be slower coming along in development?

MARC LIPSITCH: Yeah, I mean I think a rapid test would of course be very valuable if it was accurate and the value would be that you could perhaps, you know, screen people as they walk into work or as they walk into school or something and have the answer soon enough to act on that day. I’m not an expert on these kinds of tests. I do know just from my own experience that when I needed a flu test, I looked it up to see what kinds of tests were available and went to the local pharmacy because I was traveling and didn’t have an easy way to get a PCR test. And if I remember correctly, the sensitivity of the flu antigen test is something like 70%, so that’s not particularly great.

And I don’t know the technical challenges of a new test for this particular virus, but the trade off I think would be the sensitivity compared to the speed. So I think, indeed, it’s an important thing to try to develop and there’s now I think money to do that. But PCR is also an important part of it and I think we shouldn’t lose sight of the need to scale up the ability to do PCR testing or some kind of virus testing that that looks at the genetic material. It could be sequence based testing or other approaches.
I think one thing that’s really clear about testing is the type of testing you need is dependent on what it is you’re trying to do. So actually, maybe I’ll share my screen. I made a slide for a talk I’m giving tomorrow. Let’s see if it lets me share it.

MODERATOR: It should.

MARC LIPSITCH: Yeah. Okay, so, I made this slide for a talk tomorrow and it’s pretty much out of my head. This is not like the WHO official target product profile, but there is a process by which, when you’re trying to develop a test or a vaccine or a medical countermeasure, you make a list of the properties you would like it to have and then whether which ones are those are essential, and which ones of those are preferred but not essential. And so just to start a discussion – this is again not nothing official, but this is just from my head – this is a list of the things you might care about in a viral test. You want low false negative rates. You want high false positive rates. You want speed. You want to be able to do lots of tests per day and get the results back to the individual. You want to use few reagents, you want to use few swabs. You want to, you might want it to be quantitative to say how much virus and not just whether it’s there or not. And you might want it to be detecting infectious virus and not just the genetic material or the antigen.

And so, this grid sort of suggests in which cases you would care more about some of these properties than others. So, three of the reasons why you use viral testing or for diagnosing individuals the first time and, you know, just detecting who’s got it and who doesn’t. And then maybe for isolation and quarantine purposes and other purposes for patient care during the course of the illness. And a third purposes for surveillance to figure out just generally how much infection is in the community.

And without going through all these grid squares, you can see that that you need more of these things for initial diagnosis because you don’t really have time to wait all day or several days to find out if you’re trying to do contact tracing, or exclude someone from work, or the like. For patient care you need many of these things and for surveillance, you can put up with some weaknesses. So, it’s not so much we need one kind of test or another, it’s really for certain types of applications, we need, we need tests with particular characteristics and speed is of course most critical for the test and trace or for other case-based interventions. And so, I think what Dr. Birx was probably referring to is that for that purpose, it would really be valuable to have a speedy test, but it also has to be sensitive and specific.

MODERATOR: Did you have a follow up to that question?

Q: Yeah, it sounds like – and actually, from your thoughts, I mean, it sounds like there’s going to be a testing regimen or flow chart or that businesses or, you know, institutions will have to decide, you know, the right test at the right time. Is that what we should be expecting to see?

MARC LIPSITCH: Yeah, I think so. And I would say I haven’t yet heard anyone who has an algorithm for, you know, if you’re going to open up your business, who do you test; how often do you test; what do you do with the positive; what do you do with the negative. What do you do with the context of the positives, I mean, what you do with the negative is pretty clear, so what you do with the positives. What do you do with their contacts is still being worked through by everybody that I’ve talked to, I don’t think anybody’s completely got that one sorted out yet.

Q: Thank you.

MARC LIPSITCH: It needs to be, obviously, for the back to work question.

MODERATOR: Next question.

Q: Yeah, so this is not a new question, but since it remains unknown. I just would like your latest take on the total number of cases now, or rather the multiplier or the range of multipliers atop the number of diagnostic cases.

MARC LIPSITCH: Yeah. I think it’s so local that it’s really hard to give a number because we don’t know the local number in almost any place in the world. I think the numbers coming out from a place like New York are probably very reliable, the most reliable. They’re not perfectly reliable because the major problem with these numbers is – with these serologically based numbers, which I think is the ultimate answer – is that the serologically based numbers suffer badly when they’re low, meaning they’re very inaccurate when they’re low because test specificity limitations can totally swamp any signal. So, I think the Santa Clara numbers for example are almost not quite meaningless, but very close to meaningless because almost all of it could be false positives, maybe even all of it, although we know there were 1000 cases, so it couldn’t be all that but nothing about the data excludes the possibility that it’s almost all false positives.

In New York, it’s clear the test is not giving 15 or 20% false positives. And so, the numbers are much more reliable. On the other hand, New York also has much better testing of other kinds. And so, the multiplayer might be lower in New York than it is somewhere else. So, I think it’s really only meaningful at this point to give numbers that are local. And then as we build up a bunch of case studies of local numbers in places with high prevalence, then we might be able to make generalizations about the country.

Q: Okay, thank you. Second question. There’s a lot of optimism now about at least two vaccine efforts. We know that vaccines are notoriously difficult to sort of wrestled to the ground and things can go – something that seems promising can completely flop in clinical trials. So, what’s your sense whether we might actually have a vaccine in mass production come this fall?

MARC LIPSITCH: I think like many other people it’s hopeful, but cautiously, but only cautiously hopeful.I mean, I think, as you say, there’s the Moderna and the Oxford vaccines that are the furthest ahead and they’re – you know, I think if I had to put a guess on it, I would maybe give it 20% chance that, maybe 20% for each of them being more optimistic, that they get through to that point with no problems. And that’s sort of based on the fact that about a third to a half of vaccines in the old-fashioned progression from Phase two to Phase three to approval fail near the late stage. And these are less tested than those other ones. So, you know, being very, very rough, I would say maybe together there’s a one in three chance that that one of them is good enough. Maybe one in two, if we’re really lucky.

And then, the production has to be scaled up, and I think one of the really challenging topics will be how the distribution of that happens because there’s going to be an awful lot of both political and economic pressures on who gets access to those vaccines. But, you know that number is probably about worth what you paid for it, but I think that’s sort of where I am right now in my level of optimism. And, you know, they’re both very advanced organizations that are doing lots of great work. And I really hope that that the optimism side is the one that comes out, but we will have to see.

Q: Thank you. Appreciate that.

MODERATOR: Next question.

Q: Looking at the article or at the issue of how in Colorado, our state just descended this past weekend. So, we’re looking back at how effective social distancing and some of those measures were in the first weeks of the pandemic. And apologies if you’ve already addressed this, but I mean, is there any sense or any data and analytics that have shown how effective these metrics have been elsewhere in the nation? Are there any good metrics that you like to rely on for that type of analysis and how much of a challenge, frankly, has the lack of testing been in trying to really get some good data on that? I ask because there’s been a lot of skepticism out here on the cost benefit of those measures and so just some outside perspective on that.

MARC LIPSITCH: I think that every place that has imposed them almost has seen a slowing of the growth of cases and some places – Austria is just some data that I happened to look at yesterday has seen a real decline in cases. Other places seem to be flat, but not going down like Massachusetts, as we discussed. So, I think the one thing that we have learned is that intense social distancing can slow the spread of this virus.

We saw that comparing what happened in Wuhan to what happened in other parts of China. We saw that comparing across countries and we have seen that, because in the United States, because most places or many places have measurably slowed transmission and limited testing does reduce your ability to quantify that, but it doesn’t – but to the extent that testing has been getting better over time. If anything, you should be detecting a higher fraction of cases and therefore it should look worse than it is.

So, the fact that it’s flattened out in many places seems like pretty compelling evidence that it works. So that’s kind of my general take. I don’t know if you’d like further discussion.

Q: Appreciate it. Thanks. The other thing is, in terms of modeling, a lot of – there’s been a lot of talk about, you know, just the different models and kind of in terms of how we stack up with initial predictions of the virus. I mean, are there any certain models that you like to clue on a little bit more than others and some that you’ve found to be a little more – are not working?

MARC LIPSITCH: It’s a long, long answer. I mean, I think I’ve been saying for a while that people make sometimes make analogies between weather prediction and an infectious disease prediction and there are some such analogies. The key difference in my mind is that we can’t change the weather. Even with a Sharpie, we can’t change the weather as much as we would sometimes like to. And so, any prediction of what the weather is going to do is a prediction of what the weather is going to do that can be tested and nobody has the excuse, ‘well, I was assuming that nobody was going to stay home or everybody was going to stay home’ because that doesn’t affect the weather.

With infectious diseases, the predictions cause people’s behavior to change, other things cause people’s behavior to change me and the infectious disease models really are if-then statements, they’re not predictions. They’re predictions that if such and such happens, then such and such is the likely range of outcomes. And so I would distinguish more between – I mean, I think there are some models that are more reliable than others but really crucial, and sort of swamping the details of most of those models is the time frame over which they’re looking and the the assumptions about what happens.

So, I think there are some questions about, for example, how many undetected cases you think there are and how much that leads to build up of immunity faster than we expect. And those are all limitations of almost every model but, the more important aspect to me is that just to take the Imperial model, which is one of the better tested and better-known models. Most of the criticism has been that that model has been, that it under the unmitigated scenario, projected a very large number of deaths and that, in fact, we’ve had a small number compared to that of deaths, but of course those deaths have been under a mitigated scenario. And that’s exactly what the model predicted.

So I think I put more stock in models that are a) mechanistic, meaning they take into account infectious disease transmission dynamics, which is essentially all the prominent models, except for the IHME one, and b) are calibrated to data and try to account for the uncertainties in that data appropriately and to, sort of, tie themselves to the most reliable of those data.

And then the last thing I would say, and one reason that we have not as a group, tried to make detailed long-term projections about numbers of cases and numbers of deaths and those kinds of thing, is that very many models can fit the short-term trend in the period when there’s a relatively large number of susceptible people. The large differences in predictions in models come from how many cases those models believe have happened that we didn’t detect and therefore, how fast is herd immunity building up, what is the peak number of cases, and what is the timing of that. And so one of the challenges is that until we get better serology, there are lots of models that fit the data in the short-term, and some of them are going to be way more right than others in the long-term.

Q: Thank you so much. I appreciate it.

MODERATOR: Okay, I think that looks like our last question. Dr. Lipsitch, do you have any final words before we end the call?

MARC LIPSITCH: No, thanks, thanks everybody. It was good discussion and I look forward to doing it again sometime soon.

This concludes the April 30 press conference.

Barry Bloom, professor of immunology and infectious diseases and former dean of the school, and Bill Hanage, associate professor of epidemiology (April 29, 2020)

Michael Mina, assistant professor of epidemiology (April 28, 2020)

Michael Mina, assistant professor of epidemiology (April 27, 2020)