Feb 9, 2012 (CIDRAP News) – The fatality rate for officially confirmed human cases of H5N1 avian influenza infection is a stunningly high 59% (345 deaths in 584 cases). But the current controversy over publishing data about transmissible H5N1 viruses has revived a debate about whether the virus is as lethal as those numbers say.
Some proponents of publishing the full details of two studies involving H5N1 viruses that spread by airborne droplets in ferrets say the true case-fatality proportion is probably much lower, because, they suspect, many mild or asymptomatic cases have gone undetected.
If the real number of infections—the denominator—is much higher, the percentage of fatal cases drops. Some have suggested that the real fatality rate is "orders of magnitude" lower.
This argument, however, swims against the tide of scientific findings.
The primary way to detect asymptomatic or subclinical cases is to conduct seroprevalence studies—to look for H5N1 antibodies in people who weren't sick but may have been exposed to the virus, such as contacts of confirmed case-patients, poultry cullers, or residents of a village where poultry outbreaks occurred.
This has been done a number of times, and in nearly all the studies conducted since 2003, the researchers found very few people who had H5N1 antibodies, if any. For example, in a systematic review published in January 2011, Maria D. Van Kerkhove, PhD, and colleagues listed 20 seroprevalence studies, and in the 17 studies conducted since 2003, the proportion of seropostiive subjects ranged from 0 to 2.8%, with most of them reporting none.
However, experts say the data are clouded by several uncertainties. One is that researchers differ in their methods and in the antibody titer levels they use to define a positive finding. Another problem is that it's not clear how long H5N1 antibodies persist in the blood. If the antibodies wane with time, a person who is tested months to years after exposure to the virus may test negative even though he or she might have carried antibodies in the past.
Still another difficulty is that if someone had an H5N1 infection and is later tested with an assay based on a different clade (strain) of the virus, the test might not detect the antibodies, experts say.
"There are many, many uncertainties," said Tim Uyeki, MD, MPH, MPP, deputy chief for science in the Epidemiology Branch of the Influenza Division at the US Centers for Disease Control and Prevention (CDC), who has been involved in many clinical studies of H5N1 patients and H5N1 epidemiologic investigations in various countries.
Even with all the unanswered questions, though, most experts seem to think it's unlikely that the number of undetected H5N1 infections is very large.
Few positives found
Whether they tested poultry workers, healthcare workers, or patients' family members, most researchers since 2003 have found few people with antibody levels suggesting they had unrecognized infections.
The bulk of the seroprevalence studies have focused on people with possible occupational exposures. For example, a 2006 study in China's Guangdong province showed that only 1 of 110 poultry workers was seropositive for H5N1. In 2009, testing of 97 firemen, government workers, and veterinarians who had responded to an H5N1 outbreak on Ruegen Island in Germany revealed none with evidence of infection. And in 2007, none of 500 poultry workers and cullers on Vietnamese farms that had H5N1 outbreaks in 2004-05 tested positive.
The picture has been much the same for healthcare workers. In Thailand and Vietnam in 2004, three research teams tested healthcare workers, nearly all of whom had contact with confirmed H5N1 patients. None of the 168 workers tested positive.
The hunt for undetected cases also has included people without known occupational exposures. One team tested 351 Cambodians living in villages where two human cases occurred in 2005; they found no positives, according to their 2006 report. The same group conducted a similar study in villages where H5N1 cases occurred in 2006. This time 7 of 674 people (1%) tested positive.
Another Cambodian study by different investigators in 2007 showed that 18 (2.6%) of 700 people living in a village where an H5N1 case had occurred were seropositive, according to their 2010 report in the Journal of Clinical Virology.
Some contrasting findings
A curious contrast to this pattern of findings, however, is provided by studies conducted after the first human outbreak of H5N1 cases, which occurred in Hong Kong in 1997 and involved 18 cases with 6 deaths. Subsequent serologic studies found considerably higher proportions of people who carried H5N1 antibodies though they had not been sick.
For example, testing of 51 household and social contacts of Hong Kong H5N1 patients found 6 (12%) who were seropositive. In another study, 9 (3%) of 293 government workers involved in the outbreak response were found to be seropositive, and the researchers estimated that 10% of poultry workers also had antibodies. In still another investigation, 8 (38%) of 21 healthcare workers who were exposed to H5N1 patients tested positive.
Researchers have suggested that the apparent higher rate of unrecognized infections in the Hong Kong outbreak may be explained by some genetic difference between the 1997 H5N1 strain and more recent ones. In their review of seroprevalence studies, Van Kerkhove and colleagues wrote, "The higher rates of seropositivity in the studies following he 1997 outbreak may reflect the genetic differences in the viruses circulating now compared to the 1997 virus, which may have been more adaptable to human infection."
While the Hong Kong studies may represent a unique situation or strain, there is one more recent study that also contrasts with the general run of findings. A team of researchers from the United States and Thailand in 2008 tested 800 rural Thais living in an area that had numerous poultry outbreaks in the preceding years.
They found that 45 (5.6%) of the participants were seropositive for a 2005 strain of H5N1 found in Thailand, and 28 (3.5%) were positive for a 2006 strain. Positive results were more common in people older than 60 and were not statistically associated with exposure to poultry.
The study was published in the Oct 15, 2011, issue of Clinical Infectious Diseases, with Benjawan P. Khuntirat of the US Armed Forces Research Institute of Medical Sciences, Bangkok, as first author.
However, others say these findings must be taken with caution, because the researchers set a very low bar for a positive test: an antibody titer of just 1:10.
"Unfortunately I think there are some laboratory methodological issues with these results," said Van Kerkhove, who noted that all the studies she included in her review used a titer of 1:40 (by microneutralization or hemagglutination inhibition, or HI) as the criterion for seropositivity. She works in the Medical Research Council Centre for Outbreak Analysis and Modelling in the Department of Infectious Disease Epidemiology at Imperial College London.
Likewise, Uyeki commented, "I would say it's difficult to interpret these data because such a low antibody titer was used for the inclusion criteria." He said a positive finding with a titer of 1:10 may represent cross-reactive antibodies to human influenza A viruses, nonspecific cross-reactivity, or a low H5N1 titer that had declined over time—or it might mean nothing.
The study exemplifies the problem of different methods and criteria used by different investigators and labs, making findings hard to interpret.
"There are a number of H5N1 antibody studies that have been conducted in several countries, and they've been conducted in different populations with different sampling methods and different lab methodologies," said Uyeki. "And there's no one standardized approach. If you took the same serum specimen and tested it in different laboratories that did not have expertise in H5N1 serological testing methods, you might get very different results."
Van Kerkhove seconded the need for more consistent approaches. Referring to the 2008 Cambodia study, she commented in an e-mail, "We should encourage more studies like this to be conducted so that we have a better understanding of the extent of asymptomatic H5N1 (and other high path AI strains), but we should encourage standard methodology (both from the epidemiologic and laboratory sides) so that results can be compared appropriately."
She noted that the European Centre for Disease Prevention and Control (ECDC) convened a second international workshop meeting last December with the aim of achieving greater consistency in both epidemiologic and laboratory methods in seroepidemiologic studies of influenza.
The World Health Organization (WHO) has published criteria for detecting H5N1 antibodies, specifying an antibody titer of 1:80 or higher with one type of test and confirmation by a comparable result with a different assay. For research, as opposed to clinical testing, the agency recommends using a microneutralization (MN) assay. That test requires the use of live H5N1 virus and therefore can be used only in a biosafety level 3 lab.
CDC scientists developed MN for highly pathogenic H5N1 after determining that HI, the standard serologic test for flu viruses in humans, was less sensitive for detecting avian flu viruses, according to a 1999 report in the Journal of Clinical Microbiology (see link below).
Noting the difficulties involved in serologic testing, the WHO statement says, "In general, standard panels of reagents for H5N1 and other novel strains are not widely available and results among the laboratories performing these tests vary widely."
In the 20 studies reviewed by Van Kerkhove and colleagues, 14 used MN assay as the primary test, while four used HI. In one study the authors used MN or HI, and in another they used MN or ELISA (enzyme-linked immunosorbent assay).
Another difficulty in interpreting serologic findings is the uncertainty about how long H5N1 antibodies persist after infection.
"Where there's always been a bit of controversy is that people say most of the antibody work has been done well after there has been exposure," said Angus Nicoll, MB, CBE, head of the influenza program at the ECDC in Stockholm and honorary professor at the London School of Hygiene and Tropical Medicine. "Could there have been mild infections with a relatively transient antibody response or maybe no response at all?"
"I think what we'd like is to have some quick studies around some fresh cases, to make sure there aren't some people who produce antibodies for a while and then they wane," he added.
Uyeki agreed that there are gaps in the understanding of the duration of immune response to H5N1. "A person who is confirmed to have an H5N1 virus infection and develops severe illness but eventually recovers may develop a very robust antibody response, and that person may have detectable antibodies for 1 or 2 years or longer," he said.
"But for persons with asymptomatic infection or who were infected and experienced mild illness, the duration and kinetics of the H5NI virus antibody response are less well understood," Uyeki added. "Furthermore, there are probably differences by age."
"For example, I saw a young child who had experienced fever and upper respiratory tract illness and recovered quickly," a case that was included in a New England Journal of Medicine report in 2006, he said. "We were able to test serum samples collected during his illness and 3½ months later, which demonstrated a low but significant increase in H5N1 virus antibody titer." If the child had been sampled 6 months or a year after his illness, he might not have had detectable H5N1 antibodies, Uyeki said, adding that some infected persons might not mount a detectable antibody response at all.
If a person is tested for H5N1 a year after his or her likely exposure, and the result is negative, it doesn’t conclusively rule out infection, he said, adding, "Optimally you need to sample the study population about 3 to 4 weeks after the exposure to allow time for an antibody response, and ideally at multiple time points after exposure."
Jeff Bender, DVM, MS, director of the Center for Animal Health and Food Safety at the University of Minnesota in St. Paul, agreed. "There have been a few studies that have looked at how quickly antibodies die off or how long they persist," he said. "Clearly that's a potential problem in trying to interpret that data. . . . . That's why some folks use lower titers, to pick that up. But that could lead to false-positives."
In a study published in PLoS One in May 2010 (see link below), researchers in Southeast Asia showed that people who had severe H5N1 illness had a stronger and longer-lasting antibody response than did people who had asymptomatic infections. At several time points, they tested serum samples from 11 severely ill Vietnamese patients and from 31 Cambodians who had asymptomatic infections.
The mean MN antibody titers in the sick patients were 1:540 at 1 to 2 months and 1:173 at 1 to 12 months after illness, compared with 1:149 and 1:62 in the asymptomatic individuals. Also, they found that positive antibody titers persisted more than 2 years in those who had been sick, whereas 10 of 23 samples from the asymptomatic people tested negative 10 to 11 months after exposure to the virus.
Still another complication in seroprevalence studies is the ever-growing genetic diversity of H5N1 viruses.
"There are now 20 specific [H5N1] clades that have been identified," said Uyeki. "If you are in a country where, say, a clade 2.1 virus has been circulating among poultry with sporadic transmission to humans, if you use a clade 1 virus in the serologic assay, you may get invalid results. At the same time, if you use a virus from 4 to 5 years ago, it can tell you one thing, but you need to look at more recently circulating viruses. H5N1 viruses continue to evolve among poultry and are a moving target."
Seeing through the fog
Despite all the fog around the numbers, experts like Nicoll, Uyeki, and Van Kerkhove say it's unlikely that the few asymptomatic H5N1 cases suggested by the seroprevalence studies represent only the tip of a huge iceberg.
"It's been a perpetual thing that people will say, 'There must be mild cases out there.' When you look through the seroprevalence data, it's pretty unrewarding for that," said Nicoll.
For years people have been saying that there must be many asymptomatic cases and that scientists just weren't looking hard enough for them, "but that just hasn't been borne out by experience," he added.
"Even if you found some, you'd have to have a lot of them to bring the current H5N1 infection-fatality rate down even to what is estimated for the 1918 pandemic," Nicoll said.
Uyeki generally concurred, commenting, "Having said all the caveats about these studies, none of them suggest large proportions of asymptomatically infected people to date. But this could change as H5N1 viruses evolve, and they clearly are evolving, no question about that."
What about the numerator?
Uyeki said there is another, seldom-mentioned question involved in the issue: the accuracy of the official number of H5N1 deaths—the numerator of the case-fatality proportion.
"From my perspective the question is what is the numerator and what is the denominator?" he said. "Are all fatal cases being identified? And I would argue that the answer is no. Both the numerator and the denominator are likely underestimated, and the question is by how much. We really don't know."
Even in countries where H5N1 is endemic, H5N1 surveillance could be missing some hospitalized patients with severe pneumonia who may have had contact with sick poultry, if H5N1 is not suspected and not tested for,he said. A larger number of deaths, of course, would raise the case-fatality proportion.
Taken all together, the data from surveillance, epidemiologic investigations of confirmed H5N1 cases, and seroprevalence studies suggest that some infections have been missed and that the denominator is larger than the number listed by the WHO, Uyeki said. He noted that clinically mild H5N1 cases occur, most commonly in children.
"At the same time, fatal cases reported are probably an underestimate of those that have occurred," he said. "So the case-fatality proportion is likely to be lower, but it's not really clear that it's orders of magnitude lower."
And as various experts have pointed out in the current debate over publication of the findings on airborne-transmissible H5N1 viruses, the H5N1 case-fatality proportion could be a small fraction of the current 59% and still be much higher than the 2.5% level seen in the calamitous 1918 flu pandemic.
January 2011 Van Kerkhove review of risk factors for H5N1 transmission
WHO recommendations and procedures for detecting H5N1 virus in human specimens
1999 J Clin Microbiology study abstract noting low sensitivity of HI to H5N1 antibodies
May 2010 PLoS One report on the duration of antibodies in those infected with H5N1
Information on December 2011 ECDC meeting on global flu seroepidemiology