Sexually Transmitted Diseases  >  Screening Tests  >  Page 2

Screening Tests To Detect Chlamydia trachomatis and Neisseria gonorrhoeae Infections - 2002

Previous Page

Table of Contents

Next Page

Section on this page:

• C. trachomatis and N. gonorrhoeae Test Performance When Used for Screening

C. trachomatis and N. gonorrhoeae Test Performance When Used for Screening

Numerous evaluations of C. trachomatis and N. gonorrhoeae tests when used for screening have been published. NAAT sensitivities have consistently exceeded the sensitivities of non-NAATs. However, the majority of published estimates of test performance are uncertain because of probable bias from 1) misclassification of study subjects' infection status by reference standard tests; 2) using reference standard tests that are more suitable for determining whether a study subject is infected at one anatomic site (e.g., the endocervix) versus determining whether the study subject is infected at any anatomic site; and 3) predominance of studies of female and symptomatic male patients from STD clinics who might not be representative of populations who would be targeted for screening. Estimates of the differences in the performance of NAATs and non-NAATs are also uncertain because only a limited number of evaluations have been conducted that evaluated the two sets of tests by using the same set of study subjects (i.e., a head-to-head study).

Identifying a suitable standard for classifying study subjects with respect to true infection status has been difficult. Customary reference standards (e.g., culture) are highly specific but lack sensitivity. When such standards are used, evaluated test specificity is probably underestimated. Truly infected subjects who are misclassified as uninfected by the reference standard are more likely to be determined positive (and thus, inappropriately considered falsely positive) by the evaluated test than are the uninfected subjects. Misclassification of evaluated test results as false-positives occurs more frequently if the evaluated test (e.g., a NAAT) is more sensitive than the reference test (e.g., culture). The majority of evaluators of C. trachomatis and N. gonorrhoeae tests have attempted to address this problem by using a discrepant analysis procedure. This procedure has been strongly contested by statisticians and clinical epidemiologists because reference tests are applied dependent upon evaluated test results. This dependency contributes a positive bias to estimates of sensitivity and does not eliminate ambiguity regarding specificity estimates (70--75).

Identifying a reference standard that accurately classifies study subjects who can be infected at multiple anatomic sites is also problematic, chiefly for women. The majority of test evaluations for women use an endocervical reference standard only. Estimates of evaluated test sensitivity are probably higher and specificity lower, when the reference standard for infection is based on a positive result from a single anatomic site (e.g., an endocervical swab) rather than from multiple potentially infected sites (e.g., an endocervical swab or urine specimens) (54 ,76 ,77). The former standard, which produces estimates by using a specimen from one anatomic site, can be considered a specimen standard and addresses the concern of test performance regarding identification of infection at that site. The latter standard can be considered a patient standard.

The following sections summarize C. trachomatis and N. gonorrhoeae test performance on the basis of recent test evaluation studies, conducted without discrepant analysis, and consultation with specialists in test evaluation.

C. trachomatis Tests for Screening Women and Men

Sensitivity

Reviews of screening tests for C. trachomatis conclude that sensitivities of commercial NAATs exceed those of non-NAATs (14--23). A substantial proportion of the published evaluations of NAATs cited in these reviews have relied on discrepant analysis. However, certain studies permit the calculation and comparison of NAAT and non-NAAT sensitivities by using culture as an independent reference standard. A five-center study reported sensitivities of NAATs for endocervical specimens that exceeded the sensitivity of a nonamplified nucleic acid hybridization test by 19.7% (95% confidence interval [CI] = 12.9%--26.6%) for LCR and 12.4% ( 95% CI = 2.1%--22.7%) for PCR (78). The sensitivities of LCR and PCR were slightly lower when performed on urine specimens than on endocervical specimens (83.4% versus 91.4% and 79.5% versus 84.0% , respectively). The sensitivity of the nonamplified hybridization test (71.6%) was similar to that reported in an earlier study (75.3%) that compared the hybridization test, three EIAs, and a DFA performed on endocervical swab specimens by using culture as the reference standard (79). Sensitivities of the tests evaluated in that study were 61.9%--75.3%.

Culture, a NAAT (LCR), and the nonamplified hybridization test have also been compared by using an independent reference standard (78). For this evaluation, the reference standard was a positive PCR performed on an endocervical or urine specimen. For endocervical swab specimens, the sensitivity of LCR (85.5%) exceeded that of culture (74.7%) by 10.8% and that of the hybridization test (61.9%) by 23.6%. For the urine specimen, the sensitivity of LCR was 80.8%.

A limited number of studies have used an independent reference standard to compare the sensitivities of tests for detection of C. trachomatis infection in asymptomatic men. One five-center study compared LCR and PCR performed on urine from asymptomatic males by using culture of intraurethral swab specimens as the independent reference standard (80). The sensitivities of LCR (84.4%) and PCR (85.4%) were similar.

The majority of C. trachomatis test evaluation specialists consulted for these guidelines believe that the sensitivities of the point-of-care tests for C. trachomatis are substantially less than the sensitivities of tests with longer processing times. However, published studies do not provide a basis for specifying the magnitude of the difference.

Specificity and Positive Predictive Value

At the prevalences of C. trachomatis infection typical of populations who are screened (e.g., 2%--10%), the specificity of the screening test and the infection prevalence are key variables because they strongly influence the proportion of positive test results that reflect infection. The term used for this proportion is the positive predictive value (PPV) (81,82).^§ At lower prevalences, a positive screening test might need to be followed by an additional test to ensure an adequate PPV.

Published evaluations of NAATs that have used alternative target NAATs^¶ to perform discrepant analysis have reported increased specificities (e.g., exceeding 99.0%--99.5%); such reports have addressed NAATs for detection of C. trachomatis in endocervical specimens, male urethral specimens, and male or female urine. However, statisticians and clinical epidemiologists have criticized these estimation procedures also. Published studies or studies described in package inserts that have used DFA to perform discrepant analysis or have used culture as the standard without performing discrepant analysis have reported specificities of 94.1%--99.5% (75). In 1993, CDC reported a similar range of specificities for nonamplified nucleic acid probe and EIA tests to detect C. trachomatis (31). This report includes a graph of PPVs associated with a test with a sensitivity of 85% across a range of test specificities and prevalences typical of those reported for C. trachomatis screening applications (Figure).

All positive tests should be considered presumptive evidence of infection. Regardless of screening setting, screening location, patient population, and patient characteristics, a false-positive test result for C. trachomatis can have adverse medical, social, and psychological impacts for a patient. In such a case, consideration should be given to increasing specificity by performing an additional test after a positive screening test and requiring that both the screening test and additional test be positive to make a diagnosis of C. trachomatis infection (83). PPV of the diagnosis will be increased unless the screening and additional tests are falsely positive for the same reasons (e.g., both are mislabeled or both cross-react with a nonchlamydial organism). However, an additional test does not fully resolve all concerns because it might provide a false-negative result. Because therapy for C. trachomatis is safe and should not be delayed, therapy can be offered while awaiting additional test results or even if the additional test is negative. Patients with positive screening test results require counseling regarding both the risks of delaying therapy and the possibility of a false-positive test result. Consideration should be given to routine additional testing for persons with positive C. trachomatis screening tests when risk factor information or actual surveys indicate that the prevalence is low, resulting in an inadequate PPV (e.g., <90%) (see Additional Considerations in Selecting a Screening Test and also, Methods To Enhance Performance or Reduce Costs).

N. gonorrhoeae Tests for Screening Women and Men

Sensitivity

As for C. trachomatis, a substantial proportion of published evaluations of NAATs for N. gonorrhoeae have relied on discrepant analysis for calculation of test sensitivities, which might have positively biased the estimates. However, a meta-analysis of evaluations of N. gonorrhoeae summarizes sensitivity estimates for those studies for which sensitivity estimates could be based on culture as the independent reference standard (13). By using a culture standard, sensitivities were similar for an endocervical swab nucleic acid hybridization test (92.1%) and an FDA-cleared NAAT (96.7%).

Package inserts for NAATs recently cleared by FDA now include results of studies conducted by using independent reference standards without employing discrepant analysis. Two such package inserts include the results of culture, a previously FDA-cleared NAAT, and the candidate NAAT (84,85). When the independent reference standard for infection was a previously FDA-cleared NAAT positive on either endocervical swab or urine specimens, the sensitivities of endocervical swab culture were less than for the Becton Dickinson BDProbeTec (83.7% versus 90.2%) and Gen-Probe APTIMA (89.3% versus 100.0%) tests; the sensitivities of the NAATs were lower when performed on urine than on endocervical swabs (78.2% versus 90.2% and 91.7% versus 100.0%, respectively). Compared with tests for C. trachomatis, differences in sensitivities among NAATs, nucleic acid hybridization tests without nucleic acid amplification, and culture are reduced when using endocervical swabs. The exception is that culture sensitivity can decline when transport or storage conditions compromise organism viability. Gram-stain sensitivity is not presented for men because Gram stain of urethral swab smears is seldom used for screening men outside STD clinic settings, and its performance for that purpose has not been well-evaluated. Gram stain is also not widely used for screening women because of decreased sensitivity and variable specificity.

Specificity and PPV

As with C. trachomatis, test specificities and prevalences among populations screened for N. gonorrhoeae are key determinants of PPV of positive screening test results. Reported specificities of N. gonorrhoeae screening tests are similar to those for C. trachomatis. Unlike C. trachomatis tests, cross-reactivity between N. gonorrhoeae and pathogenic and nonpathogenic Neisseria has been demonstrated for certain NAATs (54 ,56 ,57). This cross-reactivity has not been reported for the Abbott LCx and Gen-Probe APTIMA and PACE 2 tests. However, nongonococcal Neisseria are infrequently recovered from the genitourinary tract. As with C. trachomatis tests, all positive screening tests should be considered presumptive evidence of infection, and routine additional testing after a positive N. gonorrhoeae screening test result would usually be indicated when screening among a low-prevalence population. Prevalences of N. gonorrhoeae are usually lower than for C. trachomatis, possibly resulting in lower PPVs. The decision to screen and conduct additional tests after a positive screening test should be made separately for the two organisms (see Additional Considerations in Selecting a Screening Test and also, Methods To Enhance Performance or Reduce Costs).

---

^§ For example, when a test with a specificity of 99% and a sensitivity of 85% is used to screen a population of 10,000 patients with a C. trachomatis prevalence of 10% (i.e., 1,000 patients have an infection), an average of 940 test results will be positive: 850 patients with a positive result will actually be infected, and 90 will not be infected (i.e., false-positives). PPV is 850/940 = 0.904. When this same test is used to screen 10,000 patients with a chlamydia prevalence of only 2% (i.e., 200 patients have an infection), an average of 268 test results will be positive: 170 patients will be infected and 98 will not. The PPV is 170/268 = 0.634.

^¶ Published evaluations of commercial NAATs have augmented reference tests with an alternative target NAAT performed by the manufacturer that employs the same amplification method as the commercial NAAT, except for substitution of a different set of primers.

Previous Page

Table of Contents

Next Page