Ebola versus influenza and some thoughts on screening

My colleague Eli Perencevich wrote an interesting blog this week in which he discusses airport screening. He points out that some of the discussion surrounding travel restrictions and Ebola are related to ideas from (and models of) epidemics of respiratory viruses, including the 2009 H1N1 pandemic. In the post he highlights some of the important differences between Ebola and influenza:

. . . Ebola is slower moving, has a much longer incubation period (especially compared to the duration of a transcontinental flight), and is not contagious before symptoms develop. What does this mean? It means that if Ebola was as infectious as influenza, millions would have already died - apocalypse. It also means that since Ebola is not transmissible during its long incubation period, it may be possible to quickly isolate patients when symptoms develop. Thus, airport screening on exit or entry could limit transmission and perhaps through early diagnosis allow Ebola infected patients to receive life saving treatment more quickly. 

Later in the post he highlights the need for mathematical model-based analysis of the impact of specific Ebola screening programs. I recommend reading the blog.

As I mentioned in a comment to the piece, in addition to incubation period, it's useful to consider the serial interval (the period between infection and transmission; sometimes also referred to as the generation interval or generation time) and basic reproduction ratio (R₀). As discussed before, estimates for R₀ for Ebola in this event are similar to estimates of R₀ for pandemic influenza events. White and Pagano estimate the serial interval for a 1995 outbreak of Ebola in Congo to be 5.4-7.6 days and the WHO Ebola Response Team estimates the serial interval for the current epidemic to be near 15 days. By comparison, estimates of the serial interval for the 2009 pandemic of influenza fall in the range of 2.5-3.0 days. Ebola has much longer serial intervals than does influenza.

What do we take away from this? One thing is that the serial interval is important for understanding the speed of spread. Perencevich observes that

. . . the first case of Ebola is thought to have occurred 307 days ago on December 6th in a two-year old boy. Since that time there have been an estimated 8,032 cases (granted these could be underestimates). If you compare a similar 307-day period for 2009 H1N1, April 12, 2009 to February 12, 2010 CDC estimated that between 42 million and 86 million cases occurred in the US with a mid-level estimate of 59 million people infected. Think about that -- 7,300 times more cases of H1N1 using the mid-level estimate during the same 307 days.

It's clear, then, that equating influenza and Ebola on the basis of R₀ alone is misleading. Thinking of R₀ as a reproductive factor for each generation of infection (at the beginning of an epidemic in a susceptible population) and the serial interval as how rapidly generations of infection occur, however, it becomes clearer that the much shorter serial interval of influenza is related the explosive emergence of influenza cases in 2009-10 relative to Ebola in 2013-14, despite the similar R₀ values. It's more complex than this in reality; Wallinga and Lipsitch present a detailed mathematical treatment of how generation intervals shape the relationship between epidemic growth rates and reproductive numbers, and Lipsitch et al illustrate, within the context of SARS, how incubation period, serial interval, and epidemic growth rate combine to produce estimates of R₀.

Another thing to ponder is that longer serial intervals can, depending on the length of the incubation period, give more time to institute control measures. On the one hand, the long serial interval relative to incubation period in the case of Ebola may suggest a higher likelihood of detecting an infectious traveler in an airport than there is for influenza. On the other hand, the extremely low incidence of Ebola in passengers must also be considered; it may not be an efficient activity to devote resources to.

I agree with Eli that mathematical models can help shed light on such questions. Perhaps such models have been published, I admit to falling behind on the mathematical epidemiology of Ebola results in the last two weeks. 

(image source: David Hartley)

Bacterial interference and the deliberate colonization of patients

Beginning in the mid-1940s and lasting until the late 1960s, the world saw a dramatic pandemic of staphylococcal infections. This post describes a curious historical episode in research aimed at controlling Staph outbreaks toward the end of that period.

One of the fundamental ideas in ecology is that, depending on the environment and properties of individuals, some types of individuals can out compete other types. When this happens, the less successful individuals can become incompletely or completely displaced. In the 1960s, the idea of microbial competition was actively applied to clinical medicine in a fascinating series of studies, which ultimately ended in tragedy. These studies investigated an idea known as "bacterial interference": the inability of a strain of a bacterium, in this case Staphylococcus aureus, to colonize a particular site of a host following deliberate colonization of that site with another strain of the bacterium.

The notion of using bacterial interference for controlling or preventing epidemics of Staph in hospital nurseries was evaluated and several trials were carried out. How this idea came about and how the studies were done is fascinating and is described in Boris, 1968 and references therein. As the nose is one of the main ecological niches of Staph aureus in humans, newborns were deliberately colonized with an apparently apathogenic strain of Staph aureus (called "strain 502A", after the phage typing scheme then in use) by swabbing the nose and the umbilical stump shortly after birth.

The results were dramatic. Clinical and epidemiological observation revealed a striking lack of staphylococcal disease in the infant study population and in their families. As Shinefield et al 1966 summarized the situation:

It has been clearly demonstrated that artificial colonization of the nasal mucosa of newborns with one strain of Staphylococcus aureus interferes with subsequent acquisition of a second strain of S aureus. This deliberate colonization of infants shortly after birth with a staphylococcal strain of low virulence (strain 502A) has been employed to protect infants from colonization and disease with virulent epidemic strains of S aureus.

The studies on children in university hospital environments were extended to children in a community hospital setting in Light et al, 1967, and found to be effective. Boris et al 1964 applied the idea to adults.

There were reservations discussed in the literature, however. An echo of that concern can be seen in an August 3, 1968, issue of the British Medical Journal, in a short report on a NEJM paper by Light et al describing observations of bacterial interference (not involving deliberate colonization) between Staph aureus and Pseudomonas. In the report, an anonymous author referred to the trials evaluating deliberate colonizations, mentioning that

Ethical objections have been raised to this procedure, but it seems no more objectionable from this standpoint than the use of living vaccines.

Unfortunately, adverse effects soon became known, including a death from infection with the 502A strain. Writing in 1972, Houck et al reported on complications associated with bacterial interference trials. A passage from the abstract describes the death due to septicemia,

An infant of a diabetic mother developed septicemia and meningitis, probably secondary to passing an umbilical vein catheter through the colonized umbilical stump. Staphylococcus aureus 502A and Escherichia coli were isolated from blood culture before death and from autopsy cultures of blood and peritoneum. A meningeal culture grew S aureus 502A. Gram-positive cocci were identified in liver, lung, heart, and meninges. 

They also noted that 

Only two (0.5%) minor 502A infections were seen in 444 spontaneously colonized infants. The benefits of S aureus 502A programs far outweigh their hazards. Disease due to the 502A strain is more frequent when the inoculum applied to the infant is large than when it is kept below 4,000 bacteria. The fatal case emphasizes that bacteria of extremely low virulence may produce serious disease in compromised hosts and that catheterization through a contaminated umbilical stump may induce bacteremia.

Although I haven't done an extensive search for bacterial interference programs after the publication of Houck et al 1972, these activities seem to have terminated after the death.

There are so many things to ponder regarding this curious episode in the 1960s, including how the one death in a few hundred patients, interpreted by Houck et al as a risk far outweighing the hazards, contrasts with current thresholds for attributable risk. Another is the remark that pathogens "of extremely low virulence may produce serious disease in compromised hosts", and how that notion is similar to the practice of avoiding live virus vaccines in recovering HSCT patients during immune system reconstitution.

Recently, Mukherjee and coworkers observed that the beneficial fungal yeast Pichia inhibits growth of pathogenic fungi, including Candida. Candida causes oral candidiasis (thrush) in immunocompromised and immunosuppressed patients. This is exciting; one of the study authors commented

One day, not only could this lead to topical treatment for thrush, but it could also lead to a formulation of therapeutics for systemic fungal infections in all immunocompromised patients . . . In addition to patients with HIV, this would also include very young patients and patients with cancer or diabetes.

I think it's important to know about the history of bacterial interference interventions so that past issues can be recognized and actively avoided in related future investigations.

(image source: Wikipedia) 

CRE outbreaks: What should we learn?

Infections with Klebsiella pneumoniae carbapenemase (KPC)-producing bacteria are associated with significant morbidity and mortality and are increasing in incidence globally. KPC- and New Delhi Metallo-beta-lactamase (NDM)- producing bacteria collectively are referred to as carbapenem-resistant Enterobacteriaceae (CRE). They are difficult to detect and treat, and thus are an important issue in hospital infection prevention.

Studies emphasize the importance of early intensification of infection control to interrupt the transmission of KPC-producing Klebsiella pneumoniae, but it is clear that infection control efforts aren't always effective at blocking transmission of the pathogen. A dramatic demonstration of this occurred in 2011 in an outbreak at the US National Institutes of Health Clinical Center. A paper by Snitkin and co-workers describes the process that unfolded in the NIH outbreak and the role that whole-genome sequencing (WGS) of isolates played in understanding how the outbreak progressed despite early implementation of infection control procedures. The study illustrates how WGS can provide evidence for unexpected transmission routes, and concludes that "integration of genomic and epidemiological data can yield actionable insights and facilitate the control of nosocomial transmission."

I think there are additional lessons from this and similar outbreaks, including that it is less than clear how, mechanistically, infection travels from host to host. The NIH staff took every intervention that could be expected to stop the spread, but those failed to break the chain of transmission. There is very limited science behind most infection prevention interventions. We don't understand in a detailed way the trips taken by pathogens as they sojourn from one host to the next, the relative probabilities of survival along the pathways, and other important facets of the contagion process.

These gaps need significant attention -- and funding for both research and education -- if we are to identify the most effective and efficient methods for control and prevention.

(image source: CDC)