|
Repeat testing of
positive samples from a low prevalence population will increase
specificity and reduce the incidence of false positive test results.
LCR testing for
gonorrhea and chlamydia in population surveys and other screenings of low
prevalence populations: coping with decreased positive predictive value.
Zenilman JM, Miller WC, Gaydos C, Rogers SM,
Turner CF.
Sexually
Transmitted Infections 2003;79:94-97
Summary:
Question
What is the predictive value for a single positive ligase chain reaction
test in a population survey in which positive specimens were retested?
Design
In a population survey that employed ligase chain reaction tests (LCR) to
diagnose gonococcal and chlamydial infections, positive predictive values
(PPV) were estimated based on a single LCR test. The theoretically
predicted PPV for a single testing was compared with the results obtained
when positive LCR tests were retested for confirmation.
Participants
Five hundred seventy-nine participants in a population based cross
sectional household survey of adults residing in Baltimore, MD (Baltimore
STD and Behavioral Survey), who were between 18 and 35 years old and
provided a urine specimen, were tested.
Description of Tests and Diagnostic
Standard
Urine specimens were tested for N. gonorrhoeae and C.
trachomatis by LCR (LCx, Abbott Laboratories, Abbott Park, IL)
following the procedures detailed by the manufacturer. All tests in the
indeterminate optical density range were retested to provide a first test
result. Specimens positive on the first LCR testing were retested using
the same urine specimen and assay. Participants were considered positive
if both the initial and confirmatory LCR tests were positive.
The PPV, expressed in a form based on
Bayes' theorem, permits the prediction of PPV across a range of
prevalences for a test with known sensitivity and specificity. Plots were
constructed of the predicted PPV across a range of plausible values for
prevalence (0.5%)-10%) and specificity (95.0%-99.9%). The PPVs were
examined at a sensitivity of 95.0% and 88.6%.
Main Outcome Measures
The theoretically derived positive predictive value of a single PCR result
was compared with the empirically observed PPV using a reference standard
of confirmed positive result.
Main Results
Twenty (3.5%) and 39 (6.7%) participants were initially positive for C.
trachomatis and N. gonorrhoeae, respectively. Repeat tests were
unavailable for two specimens that were initially N. gonorrhoeae
positive and one specimen initially positive for both organisms. Retests
yielded positive results in 17 of 19 cases of chlamydial infection and 30
of 36 gonorrhea cases, reducing the prevalence of confirmed chlamydial and
gonococcal infections to 17 (3.0%) and 30 (5.2%) of 576 participants,
respectively. The PPV of the first LCR test was 89.5% for chlamydial and
83.3% for gonococcal infections.
The theoretical PPV of a single LCR test
at 99.0% specificity and 95.0% sensitivity was 66% if the prevalence was
2%, and 32% if the prevalence was 0.5%. If the specificity increased to
99.9%, the PPV was 95.1% at 2% prevalence, and 82.6% at 0.5% prevalence.
Assuming specificities for a single LCR test to be 99.0-99.6%, the
theoretically expected PPV would be 73-87% for a single chlamydia test,
and 82-92% for a single gonococcal test. This theoretical expectation is
in agreement with the empirical results.
Authors' Conclusions
False positive results are predicted by
Bayes' theorem and are statistically unavoidable unless test specificity
is 100%. The number of false positive observed in this study using repeat
LCR testing were predicable based on test performances and the low
prevalence of infection in the population.
Source of funding: United States
National Institutes of Health; Research Triangle Institute; Clinical
Associate Physician Program of the General Clinical Research Center,
Division of Research Resources, NIH; Abbott Laboratories.
For correspondence: Jonathan Zenilman,
Infectious Diseases Division, Johns Hopkins University School of Medicine,
Baltimore, MD. E-mail address: jzenilman@jhmi.edu.
|
.gif){kind=small}