H5N1 False Negatives in Pakistan
Recombinomics Commentary 17:20
December 28, 2007
Testing done by Pakistan's National Institute of Health earlier in December identified nine possible cases, five in the family cluster, a female doctor who had treated members of the family and three poultry cullers who were unrelated to the family. A 10th possible case, another brother in the family, died without having been tested.
Some of the testing was done using a PCR or polymerase chain reaction test that looks for traces of virus in sputum samples. In other cases the Pakistan lab looked for antibodies in blood using a test which has not been validated as effective in finding H5N1 infections.
Retesting by the WHO collaborating centre for influenza in London and by experts from a U.S. naval laboratory in Cairo - known as NAMRU-3 - did not support the initial positive findings on all the suspect cases, Hayden admitted.
The negative results may be the product of degradation of specimens due to multiple freezing and thawing of the samples, he said. Other factors could have affected the results, including when in the course of infection the specimens were taken or whether suspect cases were started on antiviral drugs before a sample was taken.
The negative results may be the product of degradation of specimens that were shipped to a WHO collaborating lab in London for confirmatory testing, Hayden said. Other factors could also have affected the results, including when in the course of infection the specimens were taken or whether suspect cases were started on antiviral drugs before a sample was taken.
"We've got a host of technical issues," he said. "There are practicalities of samples breaking down over time in the transportation."
The above comments are from recent media reports in response to the WHO situation update on the H5N1 infections in Pakistan. As noted above, ten patients tested positive by labs in Pakistan, but only one was positive in testing done by WHO regional reference labs (a mobile unit shipped to Pakistan from NAMRU-3, and Mill Hill testing of samples shipped to London). Additional antibody testing will be done on serum samples, and sequencing is in progress. However, the negative data indicate that at most there will only be sequence data from the second fatal case.
The above comments also give a litany of reasons for the false negatives, but these false negatives raise concerns about past and future cases, as well as risk assessments.
The limited release of disease onset dates clearly shows sustained human to human transmission for a month, at a minimum. This chain may have stretched from the index case to a surviving brother to a deceased brother to a second deceased brother, to one or more health care workers. The exact components of the chain are still unclear because all of the onset dates have not been released, and all of the positives have not been supported by clear clinical data. However, not all H5N1 infections produce easily differentiated clinical presentations, which is why lab confirmation is important and why the false negatives increase concnerns.
Clusters are useful for delineating problems that lead to false negatives. In the cluster from Pakistan, factors for negative data include failing to test, or testing after repeated freeze and thaws or after the start of Tamiflu treatment.
The test failure for the first fatality remains unclear. Although most media reports indicate no sample was collected, at least one report quotes a health care worker as saying samples were collected from both fatal cases. It is unclear if there was no sample, or if it was lost, dropped, improperly tested, or results were lost, withheld, etc.
False negatives could have been due to sample degradation due to a failure to make aliquots of the sample resulting in multiple freeze / thaws. Freeze / thaws could also be due to packaging, shipping, or receiving issues.
There are literally hundreds of ways to generate false negatives, which is why such negatives are viewed with caution by the scientific community, but these negatives frequently drive press releases, risk assessment, and policy by others.
Unfortunately, false negatives are quite common in H5N1 testing. Most surveillance groups have yet to find H5N1 in a live wild bird. On clinical samples, negatives were eventually generated for serum samples collected from patients in northern Vietnam in 2005. In Turkey, only 12 of the 21 positives were confirmed in England, and sequences were released from only four of the positives. Many patients initially test negative, and those on Tamiflu continue to test negative because of suppression of viral RNA levels.
The assays used for H5N1 are largely confirmatory, and work best when samples are collected at the proper time and handled correctly. Virtually all human H5N1 isolates come from samples collected when the patient is near death.
Thus, in Pakistan viral isolation is limited to two patients, and samples are missing from one of the two. False negatives also lead to false transmission chain terminations, as well as failures to test contacts of the falsely negative patients.
Thus, the false negatives in Pakistan increase pandemic concerns.
Media Links
Recombinomics Presentations
Recombinomics Publications
Recombinomics Paper at Nature Precedings