394
The Biology of Sea Turtles, Vol. II
data sets. In addition to species differences, distinct normal populations may be
discriminated by differences in age, sex, season, reproductive condition, and
genetic background. For example, Bolten and Bjorndal (1992) found that among
juvenile green turtles, several plasma analytes varied significantly with body size,
whereas others such as uric acid and cholesterol differed between the sexes.
Similarly, the normal values for plasma calcium of adult female sea turtles vary
depending on their reproductive condition. As the number of samples tested
increases, the ability to find statistical significance in small differences between
means and variances also increases (Zar, 1974). These differences may or may
not be biologically relevant.
How samples were collected, transported, stored, and processed; the analysis
method and specific laboratory procedures, equipment, and reagents used; and how
well the assay was optimized and validated for the species being tested all affect
the interpretability and comparability of test results (Meyer et al., 1992; Walton,
2001a; 2001b). Values for several plasma biochemistry parameters, for example,
varied significantly when duplicate samples from loggerhead turtles were analyzed
on two different automated machines (Bolten et al., 1992). Thus, it is important for
a study that all samples be collected, handled, processed, and analyzed in the same
way, preferably in batches in the same laboratory using the same equipment and
reagents, and sometimes even analyzed by the same technician. Each laboratory
should develop its own reference ranges for each species. The issues and methodologies involved in establishment of reference ranges and validating assays are
discussed in depth by Walton (2001a; 2001b).
Reference ranges are statistical constructs, defined as the maximum and minimum
values between which a specified proportion of the population frequency distribution
will be found. Inevitably, this means that some individuals in a normal population
will fall outside the reference range by chance alone. For example, for data that have
a gaussian (normal) distribution and a reference range defined as two standard deviations above and below the mean, only about 95% of the population will fall within
the reference interval. Thus, in a sample of 100 turtles, 5 animals can be expected to
have values more extreme (either greater or less) than these limits, and yet be
completely normal, healthy individuals with respect to that parameter.
For tests that yield categorical positive or negative readouts such as serology,
microbiological culture, and polymerase chain reaction (PCR), the performance
characteristics of the test on the basis of its ability to discriminate true positive from
true negative samples (specificity and sensitivity) must be considered (Weisbroth
et al., 1998). The sensitivity of a test is the ability of the test to detect the true
positives in a population. It is that proportion of the population that is truly positive
that yields positive test results. The proportion that tested negative is false negative.
The more sensitive the test, the fewer false negatives will result. The specificity of
the test measures the ability of the test to recognize the true negatives in a population,
and is the proportion of the population that is truly negative that is detected as
negative by the test. The more specific the test, the fewer false positives will result.
When either of these values is less than 100%, the predictive value of the test (i.e.,
how much confidence can be placed in the result being true) will vary, depending
on the true prevalence of the condition in the population. Predictive value of a
1123 book.book Page 394 Monday, November 11, 2002 11:11 AM
The Biology of Sea Turtles, Vol. II
data sets. In addition to species differences, distinct normal populations may be
discriminated by differences in age, sex, season, reproductive condition, and
genetic background. For example, Bolten and Bjorndal (1992) found that among
juvenile green turtles, several plasma analytes varied significantly with body size,
whereas others such as uric acid and cholesterol differed between the sexes.
Similarly, the normal values for plasma calcium of adult female sea turtles vary
depending on their reproductive condition. As the number of samples tested
increases, the ability to find statistical significance in small differences between
means and variances also increases (Zar, 1974). These differences may or may
not be biologically relevant.
How samples were collected, transported, stored, and processed; the analysis
method and specific laboratory procedures, equipment, and reagents used; and how
well the assay was optimized and validated for the species being tested all affect
the interpretability and comparability of test results (Meyer et al., 1992; Walton,
2001a; 2001b). Values for several plasma biochemistry parameters, for example,
varied significantly when duplicate samples from loggerhead turtles were analyzed
on two different automated machines (Bolten et al., 1992). Thus, it is important for
a study that all samples be collected, handled, processed, and analyzed in the same
way, preferably in batches in the same laboratory using the same equipment and
reagents, and sometimes even analyzed by the same technician. Each laboratory
should develop its own reference ranges for each species. The issues and methodologies involved in establishment of reference ranges and validating assays are
discussed in depth by Walton (2001a; 2001b).
Reference ranges are statistical constructs, defined as the maximum and minimum
values between which a specified proportion of the population frequency distribution
will be found. Inevitably, this means that some individuals in a normal population
will fall outside the reference range by chance alone. For example, for data that have
a gaussian (normal) distribution and a reference range defined as two standard deviations above and below the mean, only about 95% of the population will fall within
the reference interval. Thus, in a sample of 100 turtles, 5 animals can be expected to
have values more extreme (either greater or less) than these limits, and yet be
completely normal, healthy individuals with respect to that parameter.
For tests that yield categorical positive or negative readouts such as serology,
microbiological culture, and polymerase chain reaction (PCR), the performance
characteristics of the test on the basis of its ability to discriminate true positive from
true negative samples (specificity and sensitivity) must be considered (Weisbroth
et al., 1998). The sensitivity of a test is the ability of the test to detect the true
positives in a population. It is that proportion of the population that is truly positive
that yields positive test results. The proportion that tested negative is false negative.
The more sensitive the test, the fewer false negatives will result. The specificity of
the test measures the ability of the test to recognize the true negatives in a population,
and is the proportion of the population that is truly negative that is detected as
negative by the test. The more specific the test, the fewer false positives will result.
When either of these values is less than 100%, the predictive value of the test (i.e.,
how much confidence can be placed in the result being true) will vary, depending
on the true prevalence of the condition in the population. Predictive value of a
1123 book.book Page 394 Monday, November 11, 2002 11:11 AM