This post covers roughly the month of December.
A. Charting the Epidemic
A1. Epidemic trajectory
Present. There is no doubt that the rapid expansion in case counts seen in September-November 2014 are now beginning to subside. This doesn’t necessarily mean a decline in case numbers – although this does still appear to be the case in Liberia – but does mean that growth is no longer exponential. This is a good thing, even though there is a long way to go. Whether this truly represents an improvement is less clear, although I have seen little suggestion that things are still as out of control as they were six weeks ago.
- Guinea continues to see a low-level epidemic in a few areas, with higher rates in and around Conakry. If we want a picture of what a slow-burning, endemic Ebola situation might look like, Guinea is looking more and more like the exemplar.
- Liberia continues to improve overall, although as with in SL and Guinea, the situation highlights the highly local nature of this disease. For example, rates in Lofa county – which suffered a terrible outbreak earlier in the year – are now way down and MSF have handed-off their treatment centre there; indeed in the past week new cases were only seen in four counties country-wide. However, new cases continue to be consistently arising in Montserrado county and Monrovia within it. The core message in Liberia, that vigilance is needed until the very last case is over, appears to be holding strong, however there remain concerns regarding fake death certificates, highlighting the potential for future flare-ups.
- Sierra Leone remains the most troubling situation. Case numbers remain high in Freetown, surrounding areas and to the north in Port Loko and Bombali. The discovery of a previously hidden epidemic in Kono lead to a lockdown of the district and a swift response from the WHO and others. Things do appear to be improving in Bo district, however. Concerns remain that preventative messaging, particularly around burials, have not been acted upon, and that the many well-meaning actors in the country are poorly coordinated to achieve change.
Future. The best estimates I’ve seen for final epidemic size (based on the admittedly imperfect data available), do seem to suggest that things are starting to come under control:
Ashleigh Tuite (@AshTuite) December 30, 2014
Tuite and Fisman, authors of the above figure, note that this downturn is a function of improvement in their model’s “control parameter”: i.e. the impact of some combination of human interventions appears to be being felt.
Modelling. There’s also an ongoing, rumbling debate about the role of modelling in the epidemic. There continue to be concerns that projections were intentionally set high, in order to gain funds – either due to need or greed. Others, for example Ian Mackay, have seen things differently, arguing that in a crisis, we make the best estimates we can and they necessarily come with uncertainty attached. In the end, I think it all comes down to how you view the unviewable – the motives of those making predictions. And in this context, there were two excellent “explainer” pieces in Science and PNAS this month, laying out how modellers approach problems and how to read their results. Andy Dobson laid out why this epidemic is hard to predict (no comparable case series for Ebola to compare it to) but that mathematical models present our only option for quantifying our uncertainty (Dobson commentary). Eric Lofgren and colleagues buttressed this argument with a step-by-step walk-through of how modellers think and why their efforts can provide useful structure in an uncertain situation (Lofgren commentary).
However, there is also an important paper out on the arXiv from Aaron King and colleagues, who highlight that fitting deterministic models has the potential to overestimate certainty in predictions (King paper). By modelling our lack of knowledge arising from the poor quality of Ebola data using stochastic methods (they use a partially observed Markov process), the authors show how we can provide meaningful descriptions of the uncertainty present in model estimates. Which might help reduce the anger towards point predictions that prove incorrect.
A2. Epidemic parameters
At this point in the epidemic, many of Ebola’s natural history parameters are well described.1 However, the parameters that rely on the interaction of infection and response are not. Primary amongst these is the case fatality rate (CFR), which continues to vary with time and space, and which is promoting a lively debate about what is the minimum level achievable – and how do we achieve it (see Treatment, below). The WHO sitrep reports a CFR of around 60% at this point for hospitalized patients, but rates of around 35-50% seem to be more common in recent reports. Indeed, Rachel Ansumana and colleagues at the Hastings ETC in Freetown have seen the mortality rate fall to 24% since early November (Asnumana letter).
The increasing number of cases recorded has also allowed for some interesting investigations of co-factors that reside within the human body. Michael Lauck and colleagues showed that in a subsample of 49 patients from Gire et al.’s May/June Sierra Leone case series, the 13 that were infected with GB virus C (also known as Pegivirus, and previously as Hepatitis G) and that risk of death was almost halved, even after adjustment for age and sex, in those co-infected. (Lauck paper). This is very preliminary data, but given previous evidence that GBV-C may be associated with better survival for HIV-infected persons, it’s a line of enquiry that may well be pursued in the future.
On another ongoing topic of interest – undetected Ebola, either at present or prior to this outbreak – I think that I have failed to previously flag a study by Randal Schoepp and colleagues from July 2014. (Schoepp paper). Studying blood submitted to Kenema hospital between 2006 and 2008 to look for Lassa fever, the authors found that 8.6% of 253 patients tested positive for Ebola. Since these individuals had a fever, this is a selected sample of the population, but suggests that a substantial proportion of infections previously thought to be other fevers in the area where Ebola emerged in 2014 may in fact have been Ebola.
Which links neatly to a study released this week by
Sam Scarpino and colleagues also used data from the Gire case series to look at clustering of social contacts. (Scarpino paper). The authors show that a best-fit network-structured state transition model requires far greater levels of triadic closure (i.e. the chance that two of my friends are friends with each other) than would be expected at random. This is hardly surprising, since we know such features are typical within close-knit groups such as those who attend funerals or care for one-another. But a nice addition to the literature that highlights the importance of linking social factors to biological ones.
On the occupational health front, a paper from Peter Kilmarx and colleagues has quantified quite how hard the healthcare professional has been hit by Ebola, but also shows that infection rates probably peaked in August. (Kilmarx paper). The authors note that this fall-off in infection may reflect the closure of many healthcare settings, or improved preventative measures – an important question if efforts to re-open non-Ebola care settings are to proceed.
And finally, a worrying new development in recent weeks has been the emergence of post-recovery health problems for some patients. I wonder if anyone has the resources to track survivors to check on this over the next few months, or even to get at what aspect of the disease/treatment process has led to these conditions.
I know the past couple of posts have been light on visual representations, and I really can’t apologize enough about that. This month, however, I can link you a Qlik Ebola widget that automatically builds from the data available on Datamarket to provide charts and dataviz on the outbreak. Lots to explore in there. For example, here’s a figure comparing case and mortality rates at the sub-national level in each of the three most-affected countries.
I can also point you towards the sterling efforts of the Centre for the Mathematical Modelling of Infectious Diseases at LSHTM. They are now putting out weekly sitreps with a wealth of data, and using that data to make epidemic need predictions (see Containment, below). Just to scratch the surface, here’s a figure of county-specific weekly case rates in Sierra Leone by type (suspected, probable, confirmed):
B. Stopping the Epidemic
One topic that arose earlier in the year was the use of Community Containment Centres (CCC) to fill the gap when there was insufficient Ebola Treatment Centre/Unit (ETC, ETU) capacity available. These CCCs would provide minimal care, but by keeping infectious individuals physically separate from their communities, might help reduce onward transmission. Concern was raised that this amounted to sub-standard care, but in a limited resource situation in which building ETCs was going to take weeks, CCCs were built. (Notably, CCCs are classically public health orientated – benefiting the population, rather than the infected individual.) They have subsequently been assessed for safety, effectiveness, and operational feasibility (results not publicly available, so far as I know). A recent modelling exercise by Adam Kucharski and colleagues showed that if individuals accessed CCCs within three days of becoming infectious, and if such facilities effectively isolated infectious patients, CCCs could shift Re below 1 (Kucharski paper). I’m not clear how necessary CCCs are now that ETC provision has been ramped up, but it’s good to know how good they need to be, in order to impact the epidemic.
And another ongoing issue: sexual transmission. As a short report this month from Karen Rogstad and Anne Tunbridge shows, no case of sexual transmission has yet been seen – although virus may be present for several months post-infection in semen (Rogstad paper). The authors highlight the need to study this in the future; another reason that following up recovered cases might be worthwhile, if/as/when we have the resources to do so.
Behaviour change. While I am quite sure that there is lots of social mobilization going on at the ground-level in SL, Liberia and Guinea, that isn’t getting much news coverage. What is getting coverage, is bans on congregating at Christmas (in Sierra Leone and Liberia) and New Year (in Guinea, but not in Liberia). Although allowances are apparently being made for religious gatherings – which may have led to an outbreak in Monrovia.
There also remains concern that burials are not being performed in a safe manner. Quite possibly due to the perfectly valid fear that engaging with the Ebola care system is associated with death. Although of course the alternative of not engaging with the system may be even more strongly associated with death. (Understanding counterfactuals is hard enough in a graduate school classroom, let alone in the midst of an epidemic.) Sierra Leone has been conducting further lockdowns and house-to-house searches.
Whether this top-down approach will have the same impact as bottom-up change remains to be seen. The bottom-up approach is also coming in for some critical consideration this month too – Clare Chandler and colleagues at the Ebola Anthropology Response Platform highlighted the biomedical, exoticising and one-size-fits-all nature of current behaviour change messages for Ebola (Chandler comment). The authors point to a more tailored messaging programme for best results.
One of the up-sides to outbreaks such as the current one is the spur it provides for “general-purpose” technologies that should carry over to many future epidemics and non-epidemics. On this front, the more engineering-style Ebola hackathons have been very useful. Recent examples include a double-layer PPE system involving magnets to remove the highest-risk outside layer and another PPE system with battery-powered cooling systems to allow for longer workperiods. Hopefully, some of these designs will be cost- and build-feasible in the near future.
Another key aspect of treating the epidemic is identifying cases. As of last week we have a faster test available, however as Laurie Garrett noted, a three-hour waiting time and the need for a lab make it a pretty poor step forwards. The need for an accurate, rapid, field-ready test remains.
And yet another key factor is how we treat patients once they are in care. A recent articles in the New York Times played up the level of debate amongst physicians about how aggressively to hydrate patients. My impression of the discussion of this article is that there is more heat than light to this, but there are certainly differing opinions over who/when to treat intravenously, with peripherally inserted central catheters or even intraosseous insertions. Given the range of medical settings in place in Guinea/SL/Liberia, I expect to see continued variation in care and in opinion in the future.
Drugs. Turning to trials of novel treatments for Ebola, we have mixed news. As a backgrounder, I really liked this clear outline of the various Ebola lifecycle entry points for treatment, and available drugs, in an article by Dr Lai Kang-Yiu published late in November (Lai paper). I’m not a huge fan of the colour scheme, but this figure from the paper lays it all out for the more visual learners amongst us:
The idea of providing infected individuals with serum – roughly, the non-water part of blood – from recovered Ebola patients has been around for months, and in use in high-income settings. A recent overview from Thomas Kriel highlighted the importance of virus-inactivating any blood collected (primarily for non-Ebola infections), as well as methods to negate the need for blood-matching. In mid-December the first trial of convalescent serum in West Africa began in Liberia.
The not-so-good side of novel treatments is this story from a treatment centre in Freetown, where the antiarrhythmic amiodarone was being used without any formal testing process. The (NHS) clinical team threatened to walk out of the Emergency-run facility , before agreement was reached to stop amiodarone’s use. I imagine that this is only one of many difficult processes that will arise in the roll-out of novel therapies.
I’ll also note this recent call for RCTs of proposed therapies in the NEJM. Given the current fluctuations in mortality rates within and between hospitals, the case for randomized trials of some description may hold more water than it did when outcomes were pretty consistent. Especially in cases where effectiveness is not clear.
No progress on in-country trials in West Africa, that I’ve heard of. But there were a couple of vaccine stories in December. First, the second Ebola vaccine in the pipeline (VSV-EBOV; originally Canadian, now NewLink/Merck) faced some hiccups during phase 1 (safety) trials, when some participants complained of joint pain and the trial was stopped early. While some commentators saw this as a Very Bad Thing, the WHO claimed that such joint pain was not atypical for vaccines. I also noted with interest the final comment by Thomas Geisbert, who was involved in developing the VSV vaccine
[VSV-EBOV] is the best vaccine virus I’ve ever seen for the kind of viruses we work with. I don’t know why it’s so good…I’d never take [the GSK vaccine]. Not with the VSV. I’d take my chances with some joint pain.
Second, a phase 1b safety trial of a more general filovirus vaccine that has been running all year in Uganda was published showing effectiveness (in generating an immune response) and safety (Kibuuka paper). These findings have already been built into the related, but more advanced, Chimpanzee adenovirus ChAd3 that passed phase 1 testing in the US last month. But researchers were happy to see that this family of vaccines had effect in African populations. I’ll chalk this up as supporting evidence.
B4. Social factors/impact
There was some recent brouhaha over Alex Kentikelenis and several other Cambridge sociologists/epidemiologists’ letter to Lancet suggesting that the IMF’s policies in recent years had led to an underfunded healthcare system, and thus increased the impact of the current Ebola outbreak. (Kentikelenis letter). The IMF responded to the say they didn’t see it that way, and had been proactive in providing funding once Ebola had emerged. Chris Blattman, a Political Science professor at Columbia suggested the original authors misunderstood the IMF’s level of influence over health policy in the affected region. Of course, the impact of macroeconomic policy is very hard to evaluate – finding a realistic counterfactual for any action is close of impossible; so it all comes back down to how you read the tealeaves. Or at least that’s how I see it…
The flip side of financial problems as a cause of the epidemic is financial problems as the fall-out. In addition the massive macroeconomic impact of the epidemic, there are lower-level effects worth noting. Within the healthcare systems, clearly everything except Ebola treatment has been greatly scaled-back during the epidemic. This was highlighted in a recent paper by Håkon Bolkan and colleagues surveying changes in inpatient admissions and surgeries at 40 facilities across Sierra Leone in September and October (Bolkan paper). Not surprisingly, as Ebola rose, everything else fell precipitously:
The impact of Ebola on the healthcare system will have knock-on effects in years to come, especially for infectious diseases. (Although in some cases the effects are being felt already – see this recent report on the side-effects and complications being caused by antimalarials provided nationwide in Sierra Leone last month.) Effects on women are likely to be particularly severe, since they are both the most likely to be healthcare providers and consumers. Outside the healthcare sector, education has been put on hold for six months generally, agriculture and trading activity has been greatly affected, while NGO work in other sectors has also come to a halt. Food insecurity is not yet widespread, but may be considerable in rural subsistence settings.
- Helen Branswell has written an overview of “lessons learned” during this outbreak. A gentle introduction to a very complex topic.
- Laurie Garrett has written four more articles for Foreign Policy from her trip to West Africa earlier in the year. As ever, these are delivered through Garrett’s own particular lens, but nonetheless worth reading for context at least.
There are also a growing number of first-person narratives from foreign responders, worth reading to see what it is like working in ETCs or in other capacities on the front line. I present these in no particular order (see also previous posts):
- Paul Pronyk, infectious disease and public health physician working in Freetown.
- John Wright, a British physician, was in Sierra Leone last month.
- African platypus, an anonymous US nurse practitioner, was in Liberia in November/December.
- Pieter Baker, a US epidemiologist working on surveillance in Sierra Leone.
- Gillian McKay, a Canadian nurse and ETC PPE trainer in Sierra Leone.
I am only too aware that all of these are out-of-region persons who have arrived for this epidemic. I would love to include more viewpoints from locals, but am not connected to them at present. I do see some one-off posts via American/European news sources, for example this report of a day in the life of a contact tracer in Monrovia. But more pointers very welcome.
The local person I read the most is Umaru Fofana, who writes for the BBC, Politico Sierra Leone and himself on Facebook. But he is a journalist, rather than a practitioner. On that front, I’d also point you towards the range of other local journalism recently highlighted by Crawford Kilian.
D. Getting involved
- For those of you with French language skills, there’s an Ebola MOOC from Unige (l’Université de Genève) and UNF3S (l’université numérique francophone des sciences du sport et de la santé), starting on January 12th.
- If you are a modeller, you may be interested in the upcoming workshop on “Modeling the Spread and Control of Ebola in West Africa“, January 22-23 in Atlanta. They’re accepting posters too, but the deadline for that is January 8th.
If you have seen something I’ve missed (or links are broken), you can reach me @harlingg. And as ever, thank you to all those on whose intellectual shoulders I am standing.
1 This is mostly for my reference, but there was a paper published this month highlighting that phylodynamic methods can estimate key epidemiological parameters using only the 74 initial outbreak cases, as opposed to the larger number needed for count-based analysis (Samuel Alizon et al, Virulence 2014).