396
The Biology of Sea Turtles, Vol. II
1993). Negative-staining electron microscopy, on the other hand, is unlikely to detect
viruses when there are fewer than 10 4 particles per microliter of sample.
When diagnostic tests such as serology are used to monitor populations to
ensure that they are free of a particular agent, interpretation of negative test results
must take into account the probability of detection (Weisbroth et al., 1998). Even
when a test is able to detect every positive animal (100% sensitive), sample sizes
must be adequate to ensure that a population is negative. The overall chances ( P )
of detecting a single positive animal will be a function of sample size ( n ) and
prevalence ( p ) described by the equation, P = [1 – (1 – p ) n ]. Thus, one can calculate
the sample size needed for a particular level of probability of detection when the
agent has a specific prevalence. For example, to have a 99% chance of detecting
even a single turtle that is positive for antibodies to the FP-associated herpesvirus
in a population that has a true prevalence of 40% requires that at least ten turtles
be tested. If the true prevalence is only 10%, at least 40 turtles must be tested for
the same degree of confidence. Presented another way, if only ten turtles are tested
in a population that has a true prevalence of 10%, the herpesvirus would have a
35% chance of going completely undetected. Thus, the more rare the disease
condition in the population, the more animals must be sampled to have a reasonable
chance of detecting it. If one accounts for lower test sensitivities, the required
sample sizes increase.
There are also several biologically important reasons why a test may fail to
detect an abnormality or disease agent. The time that the diagnostic procedure was
performed and the sample collected relative to the disease course is important. For
example, it takes a certain period of time for turtles to mount an immune response
against a pathogen. Thus, early in the course of infection, pathogen-specific antibodies may not be detected serologically. Some infectious agents replicate only
during specific stages of the disease and sometimes can be found in different tissues
at different stages. Therefore, tissue samples collected too early or too late in the
course may yield negative results. Furthermore, in severe disease under certain
circumstances, values for a particular assay that is typically a sensitive indicator of
a disease condition may be found to be within normal limits. For example, the white
blood cell count, a sensitive indicator of an active inflammatory response to infection,
may yield counts within the normal range if a turtle is losing cells from the circulation
faster than it is able to replace them. Similarly, the elevation of certain liver enzymes
in blood indicates liver cell damage, but the levels could be within normal limits in
chronic active liver disease if sufficient liver parenchyma has already been lost.
Many factors related to sample quality, preparation, storage, handling, and contamination could affect test results in either direction. For example, exposure of a
plasma sample to light degrades bilirubin, falsely lowering its measured concentration. Contamination of plasma with hemolyzed blood causes marked elevation in
several enzymes and interferes with colorimetric measurements of some analytes
(Meyer et al., 1992). Plasma samples that have been repeatedly thawed and refrozen
have decreased enzyme activities and lower specific antibody titers.
There is a significant additional problem in interpreting the biological and
clinical relevance of some tests (especially certain blood biochemistry values) for
sea turtles. Many of the analytes tested in blood biochemistry panels were selected
1123 book.book Page 396 Monday, November 11, 2002 11:11 AM
The Biology of Sea Turtles, Vol. II
1993). Negative-staining electron microscopy, on the other hand, is unlikely to detect
viruses when there are fewer than 10 4 particles per microliter of sample.
When diagnostic tests such as serology are used to monitor populations to
ensure that they are free of a particular agent, interpretation of negative test results
must take into account the probability of detection (Weisbroth et al., 1998). Even
when a test is able to detect every positive animal (100% sensitive), sample sizes
must be adequate to ensure that a population is negative. The overall chances ( P )
of detecting a single positive animal will be a function of sample size ( n ) and
prevalence ( p ) described by the equation, P = [1 – (1 – p ) n ]. Thus, one can calculate
the sample size needed for a particular level of probability of detection when the
agent has a specific prevalence. For example, to have a 99% chance of detecting
even a single turtle that is positive for antibodies to the FP-associated herpesvirus
in a population that has a true prevalence of 40% requires that at least ten turtles
be tested. If the true prevalence is only 10%, at least 40 turtles must be tested for
the same degree of confidence. Presented another way, if only ten turtles are tested
in a population that has a true prevalence of 10%, the herpesvirus would have a
35% chance of going completely undetected. Thus, the more rare the disease
condition in the population, the more animals must be sampled to have a reasonable
chance of detecting it. If one accounts for lower test sensitivities, the required
sample sizes increase.
There are also several biologically important reasons why a test may fail to
detect an abnormality or disease agent. The time that the diagnostic procedure was
performed and the sample collected relative to the disease course is important. For
example, it takes a certain period of time for turtles to mount an immune response
against a pathogen. Thus, early in the course of infection, pathogen-specific antibodies may not be detected serologically. Some infectious agents replicate only
during specific stages of the disease and sometimes can be found in different tissues
at different stages. Therefore, tissue samples collected too early or too late in the
course may yield negative results. Furthermore, in severe disease under certain
circumstances, values for a particular assay that is typically a sensitive indicator of
a disease condition may be found to be within normal limits. For example, the white
blood cell count, a sensitive indicator of an active inflammatory response to infection,
may yield counts within the normal range if a turtle is losing cells from the circulation
faster than it is able to replace them. Similarly, the elevation of certain liver enzymes
in blood indicates liver cell damage, but the levels could be within normal limits in
chronic active liver disease if sufficient liver parenchyma has already been lost.
Many factors related to sample quality, preparation, storage, handling, and contamination could affect test results in either direction. For example, exposure of a
plasma sample to light degrades bilirubin, falsely lowering its measured concentration. Contamination of plasma with hemolyzed blood causes marked elevation in
several enzymes and interferes with colorimetric measurements of some analytes
(Meyer et al., 1992). Plasma samples that have been repeatedly thawed and refrozen
have decreased enzyme activities and lower specific antibody titers.
There is a significant additional problem in interpreting the biological and
clinical relevance of some tests (especially certain blood biochemistry values) for
sea turtles. Many of the analytes tested in blood biochemistry panels were selected
1123 book.book Page 396 Monday, November 11, 2002 11:11 AM