Cervical Specimen Order and Performance Measures of Chlamydia trachomatis Diagnostic Testing
http://www.100md.com
微生物临床杂志 2005年第10期
Johns Hopkins University School of Medicine, Baltimore, Maryland
Centers for Disease Control and Prevention, Atlanta, Georgia
University of Washington, Seattle, Washington
ABSTRACT
The orders of three endocervical specimens of 3,561 women for Chlamydia trachomatis testing were randomized to determine whether test performance measures of two nucleic acid amplification tests and a DNA probe were affected by swab order. Specimen collection order did not appear to affect the diagnostic accuracy of these tests.
TEXT
Previous studies have suggested that swab order may influence the sensitivity of non-nucleic acid amplification tests (non-NAATs) in the detection of Chlamydia trachomatis from the endocervix (3-5, 7). One study reported an increased sensitivity with later swabs when using culture to test for the presence of infection (3). Hadgu reported that the sensitivity for Chlamydia trachomatis detection was higher for the first three swabs when a series of five swabs from 4,583 women for C. trachomatis testing using nonculture non-NAATs were randomized (4). In this study, our goal was to evaluate the effect of swab order on the performance characteristics of two NAATs and one nonamplified DNA probe test used in the detection of endocervical C. trachomatis infection.
Details of enrollment and specimen collection and processing can be found elsewhere (1, 6). In brief, women who presented to sexually transmitted disease clinics at the five participating centers were eligible to participate if they had an indication that a pelvic examination should be performed. Informed consent was obtained from all participants prior to their enrollment. The study was approved by the institutional review boards at each of the participating centers and at the Centers for Disease Control and Prevention.
Patients were asked to void, and the first 30 ml of urine from each patient was saved for ligase chain reaction (LCR) and PCR testing for C. trachomatis. A pelvic exam was then performed, and a series of endocervical specimens were collected. First, a swab for Neisseria gonorrhoeae culture and/or a Gram stain was obtained from all subjects; in three of the participating centers, a Pap smear was then performed on a subset of women if clinically indicated. The next three specimens were obtained for C. trachomatis testing by (i) DNA probe (PACE 2; Gen-Probe), (ii) LCR (Abbott Laboratories, Abbott Park, IL), and (iii) PCR (Roche Diagnostics Systems, Branchburg, NJ), with the sequence randomized. Following the three randomized swabs, a cytobrush specimen was collected for C. trachomatis culture from all participants. In summary, the first of the three randomized swabs for C. trachomatis testing was either the second endocervical specimen obtained if no Pap smear was performed (hence, the order of the randomized cervical specimens was 2, 3, and 4) or the third endocervical specimen obtained if a Pap smear was clinically indicated (hence, the order of the randomized cervical specimens was 3, 4, and 5) as shown in Table 1, for centers designated C, D, and E. The presence of blood on all specimens was noted, as was the presence of cervical discharge.
An infection reference standard ("gold standard") using two of three positive tests for C. trachomatis among urine LCR and PCR, and endocervical C. trachomatis culture was used in all primary analyses. The requirement that two reference tests be positive increased the reference standard's specificity, thereby avoiding the potential confounding effect of a gold standard with a high false-positive rate. We did perform similar analyses using less-specific gold standards (one of three positives and one of two positives). Although this altered the estimates of the performance characteristics of the three diagnostic tests, it did not affect the relative differences of these estimates among the different swab orders (data not shown). Standard formulae to calculate sensitivity and specificity were applied, and exact binomial 95% confidence limits are reported. The Pearson 2 test for homogeneity was used to compare independent discrete variables, and the Cuzick method (an extension of the Wilcoxon rank sum test) was used to test for statistical trends when more than two results were being evaluated and the data were independent (2). Logistic regression models were used to adjust for the potential confounding effects of testing center, order of first randomized swab, presence of blood on the collection swab, and presence of cervical mucopus on the odds of obtaining a positive C. trachomatis result for each diagnostic test studied. Generalized estimating equations for correlated data were used to calculate performance measures when repeated observations were compared (8, 9). Statistical analyses were performed using STATA v.8.1 statistical software (College Station, TX).
Of 3,903 total visits, 342 observations were excluded from this analysis due to missing or unsatisfactory C. trachomatis diagnostic test results (n = 104), missing or miscoded randomization schemes (n = 109), and more than one clinic visit by study participants (n = 129). The final number of subjects was 3,561. The mean age was 26 years (range, 14 to 69 years). The overall prevalence of C. trachomatis infection according to the reference standard was 10.8%.
The percent positive by testing site and randomization order for each of the diagnostic tests is summarized in Table 1. The percentage of unsatisfactory specimens (as a result of an error in sample collection or laboratory processing) was evaluated for each diagnostic test stratified by testing center. There was no difference in the number of unsatisfactory specimens with increasing swab order for either PCR or LCR. Overall, there was a trend for an increased number of unsatisfactory specimens with increasing swab order for the PACE 2 test. This ranged from 0.2% with swab 2 to 1.6% with swab 5 (Cuzick's test for trend, P = 0.04). However, after adjusting the data for center, the difference was no longer statistically significant.
The test performance measures of each of the randomized diagnostic tests relative to the reference standard were calculated and stratified by center (data not shown). As no effect of testing center or order of first randomized swab using logistic regression models was demonstrated, results from all testing centers were combined, and performance measures were calculated for each of the randomized tests stratified by swab order (Table 2). There were no statistically significant differences in either sensitivity or specificity by Pearson's test for homogeneity or Cuzick's test for trend for PACE 2 (Pearson's test P value, 0.8; Cuzick's test for trend P value, 0.5), LCR (Pearson's test for homogeneity P value, 1.0; Cuzick's test for trend P value, 1.0), and PCR (Pearson's test for homogeneity P value, 1.0; Cuzick's test for trend P value, 0.9) among the four randomized swabs.
Having shown no significant differences in performance measures between LCR and PCR, results for both tests were combined, and point estimates of sensitivity and specificity of the overall C. trachomatis NAATs stratified by swab order were calculated by using generalized estimating equations. The sensitivity point estimates of swabs 2, 3, 4, and 5 were 88.6%, 86.2%, 86.6%, and 92.7%, respectively, and the specificity point estimates were 96.6%, 97.3%, 98.6%, and 98.8%. None of the differences were statistically significant.
The prevalence of blood on the swabs increased with advancing specimen order, from about 20% with swab no. 2 to 80% with swab no. 5 (Cuzick's test of trend P value, <0.0001 for PACE 2, LCR, and PCR). The sensitivity and specificity estimates were not significantly affected by the presence of blood for any of the performance measures of these tests.
Our study has limitations. We do not have data on the effects of swab order comparing the first endocervical swab to later swabs. The first endocervical swab in all participants was used for Neisseria gonorrhoeae culture or Gram stain testing. It is theoretically possible that differences in performance measures may exist between a first swab and a second swab. This study has several strengths. It was conducted in various centers that recruited participants with a broad age range. The relatively large sample size allowed for precise estimates of performance measures for each individual test and for each swab order. Finally, the randomized design allowed for the minimization of bias.
Our data suggest that the order of swab collection from the endocervix does not influence the performance characteristics of the PCR, LCR, and PACE 2 tests in diagnosing C. trachomatis infections. This finding has potential implications in both clinical and research settings, where multiple endocervical specimens are often necessary to complete the diagnostic evaluation of genital tract infections and cervical dysplasia in women.
ACKNOWLEDGMENTS
This work was supported by the Centers for Disease Control and Prevention, cooperative agreement number 455 and the Association of Teachers for Preventive Medicine (K.G.G.).
We thank the staff and subjects at the five participating sites.
Use of trademark is for identification purposes only and does not imply endorsement by the U.S. Department of Health and Human Services.
REFERENCES
Black, C. M., J. Marrazzo, R. E. Johnson, E. W. Hook III, R. B. Jones, T. A. Green, J. Schachter, W. E. Stamm, G. Bolan, M. E. St. Louis, and D. H. Martin. 2002. Head-to-head multicenter comparison of DNA probe and nucleic acid amplification tests for Chlamydia trachomatis infection in women performed with an improved reference standard. J. Clin. Microbiol. 40:3757-3763.
Cuzick, J. 1985. A Wilcoxon-type test for trend. Stat. Med. 4:87-90.
Embil, J. A., H. J. Thiebaux, F. R. Manuel, L. H. Pereira, and S. W. MacDonald. 1983. Sequential cervical specimens and the isolation of Chlamydia trachomatis: factors affecting detection. Sex. Transm. Dis. 10:62-66.
Hadgu, A. 1998. A biomedical application of latent class models with random effects. Appl. Stat. 47:603-616.
Hernandez, T. J., K. L. Noller, and T. F. Smith. 1986. Detection of Chlamydia trachomatis using consecutive endocervical swabs. Prevalence in asymptomatic female adolescents and women attending a sexually transmitted disease clinic. J. Reprod. Med. 31:497-500.
Marrazzo, J. M., R. E. Johnson, T. A. Green, W. E. Stamm, J. Schachter, G. Bolan, E. W. Hook III, R. B. Jones, D. H. Martin, M. E. St. Louis, and C. M. Black. 2005. Impact of patient characteristics on performance of nucleic acid amplification tests and DNA probe for detection of Chlamydia trachomatis in women with genital infections. J. Clin. Microbiol. 43:577-584.
Munday, P. E., J. M. Carder, N. F. Hanna, and D. Taylor-Robinson. 1984. Is one swab enough to detect chlamydial infection of the cervix Br. J. Vener. Dis. 60:384-386.
Sternberg, M. R., and A. Hadgu. 2001. A GEE approach to estimating sensitivity and specificity and coverage properties of the confidence intervals. Stat. Med. 20:1529-1539.
Zeger, S. L., and K. Y. Liang. 1986. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 42:121-130.(Khalil G. Ghanem, Robert )
Centers for Disease Control and Prevention, Atlanta, Georgia
University of Washington, Seattle, Washington
ABSTRACT
The orders of three endocervical specimens of 3,561 women for Chlamydia trachomatis testing were randomized to determine whether test performance measures of two nucleic acid amplification tests and a DNA probe were affected by swab order. Specimen collection order did not appear to affect the diagnostic accuracy of these tests.
TEXT
Previous studies have suggested that swab order may influence the sensitivity of non-nucleic acid amplification tests (non-NAATs) in the detection of Chlamydia trachomatis from the endocervix (3-5, 7). One study reported an increased sensitivity with later swabs when using culture to test for the presence of infection (3). Hadgu reported that the sensitivity for Chlamydia trachomatis detection was higher for the first three swabs when a series of five swabs from 4,583 women for C. trachomatis testing using nonculture non-NAATs were randomized (4). In this study, our goal was to evaluate the effect of swab order on the performance characteristics of two NAATs and one nonamplified DNA probe test used in the detection of endocervical C. trachomatis infection.
Details of enrollment and specimen collection and processing can be found elsewhere (1, 6). In brief, women who presented to sexually transmitted disease clinics at the five participating centers were eligible to participate if they had an indication that a pelvic examination should be performed. Informed consent was obtained from all participants prior to their enrollment. The study was approved by the institutional review boards at each of the participating centers and at the Centers for Disease Control and Prevention.
Patients were asked to void, and the first 30 ml of urine from each patient was saved for ligase chain reaction (LCR) and PCR testing for C. trachomatis. A pelvic exam was then performed, and a series of endocervical specimens were collected. First, a swab for Neisseria gonorrhoeae culture and/or a Gram stain was obtained from all subjects; in three of the participating centers, a Pap smear was then performed on a subset of women if clinically indicated. The next three specimens were obtained for C. trachomatis testing by (i) DNA probe (PACE 2; Gen-Probe), (ii) LCR (Abbott Laboratories, Abbott Park, IL), and (iii) PCR (Roche Diagnostics Systems, Branchburg, NJ), with the sequence randomized. Following the three randomized swabs, a cytobrush specimen was collected for C. trachomatis culture from all participants. In summary, the first of the three randomized swabs for C. trachomatis testing was either the second endocervical specimen obtained if no Pap smear was performed (hence, the order of the randomized cervical specimens was 2, 3, and 4) or the third endocervical specimen obtained if a Pap smear was clinically indicated (hence, the order of the randomized cervical specimens was 3, 4, and 5) as shown in Table 1, for centers designated C, D, and E. The presence of blood on all specimens was noted, as was the presence of cervical discharge.
An infection reference standard ("gold standard") using two of three positive tests for C. trachomatis among urine LCR and PCR, and endocervical C. trachomatis culture was used in all primary analyses. The requirement that two reference tests be positive increased the reference standard's specificity, thereby avoiding the potential confounding effect of a gold standard with a high false-positive rate. We did perform similar analyses using less-specific gold standards (one of three positives and one of two positives). Although this altered the estimates of the performance characteristics of the three diagnostic tests, it did not affect the relative differences of these estimates among the different swab orders (data not shown). Standard formulae to calculate sensitivity and specificity were applied, and exact binomial 95% confidence limits are reported. The Pearson 2 test for homogeneity was used to compare independent discrete variables, and the Cuzick method (an extension of the Wilcoxon rank sum test) was used to test for statistical trends when more than two results were being evaluated and the data were independent (2). Logistic regression models were used to adjust for the potential confounding effects of testing center, order of first randomized swab, presence of blood on the collection swab, and presence of cervical mucopus on the odds of obtaining a positive C. trachomatis result for each diagnostic test studied. Generalized estimating equations for correlated data were used to calculate performance measures when repeated observations were compared (8, 9). Statistical analyses were performed using STATA v.8.1 statistical software (College Station, TX).
Of 3,903 total visits, 342 observations were excluded from this analysis due to missing or unsatisfactory C. trachomatis diagnostic test results (n = 104), missing or miscoded randomization schemes (n = 109), and more than one clinic visit by study participants (n = 129). The final number of subjects was 3,561. The mean age was 26 years (range, 14 to 69 years). The overall prevalence of C. trachomatis infection according to the reference standard was 10.8%.
The percent positive by testing site and randomization order for each of the diagnostic tests is summarized in Table 1. The percentage of unsatisfactory specimens (as a result of an error in sample collection or laboratory processing) was evaluated for each diagnostic test stratified by testing center. There was no difference in the number of unsatisfactory specimens with increasing swab order for either PCR or LCR. Overall, there was a trend for an increased number of unsatisfactory specimens with increasing swab order for the PACE 2 test. This ranged from 0.2% with swab 2 to 1.6% with swab 5 (Cuzick's test for trend, P = 0.04). However, after adjusting the data for center, the difference was no longer statistically significant.
The test performance measures of each of the randomized diagnostic tests relative to the reference standard were calculated and stratified by center (data not shown). As no effect of testing center or order of first randomized swab using logistic regression models was demonstrated, results from all testing centers were combined, and performance measures were calculated for each of the randomized tests stratified by swab order (Table 2). There were no statistically significant differences in either sensitivity or specificity by Pearson's test for homogeneity or Cuzick's test for trend for PACE 2 (Pearson's test P value, 0.8; Cuzick's test for trend P value, 0.5), LCR (Pearson's test for homogeneity P value, 1.0; Cuzick's test for trend P value, 1.0), and PCR (Pearson's test for homogeneity P value, 1.0; Cuzick's test for trend P value, 0.9) among the four randomized swabs.
Having shown no significant differences in performance measures between LCR and PCR, results for both tests were combined, and point estimates of sensitivity and specificity of the overall C. trachomatis NAATs stratified by swab order were calculated by using generalized estimating equations. The sensitivity point estimates of swabs 2, 3, 4, and 5 were 88.6%, 86.2%, 86.6%, and 92.7%, respectively, and the specificity point estimates were 96.6%, 97.3%, 98.6%, and 98.8%. None of the differences were statistically significant.
The prevalence of blood on the swabs increased with advancing specimen order, from about 20% with swab no. 2 to 80% with swab no. 5 (Cuzick's test of trend P value, <0.0001 for PACE 2, LCR, and PCR). The sensitivity and specificity estimates were not significantly affected by the presence of blood for any of the performance measures of these tests.
Our study has limitations. We do not have data on the effects of swab order comparing the first endocervical swab to later swabs. The first endocervical swab in all participants was used for Neisseria gonorrhoeae culture or Gram stain testing. It is theoretically possible that differences in performance measures may exist between a first swab and a second swab. This study has several strengths. It was conducted in various centers that recruited participants with a broad age range. The relatively large sample size allowed for precise estimates of performance measures for each individual test and for each swab order. Finally, the randomized design allowed for the minimization of bias.
Our data suggest that the order of swab collection from the endocervix does not influence the performance characteristics of the PCR, LCR, and PACE 2 tests in diagnosing C. trachomatis infections. This finding has potential implications in both clinical and research settings, where multiple endocervical specimens are often necessary to complete the diagnostic evaluation of genital tract infections and cervical dysplasia in women.
ACKNOWLEDGMENTS
This work was supported by the Centers for Disease Control and Prevention, cooperative agreement number 455 and the Association of Teachers for Preventive Medicine (K.G.G.).
We thank the staff and subjects at the five participating sites.
Use of trademark is for identification purposes only and does not imply endorsement by the U.S. Department of Health and Human Services.
REFERENCES
Black, C. M., J. Marrazzo, R. E. Johnson, E. W. Hook III, R. B. Jones, T. A. Green, J. Schachter, W. E. Stamm, G. Bolan, M. E. St. Louis, and D. H. Martin. 2002. Head-to-head multicenter comparison of DNA probe and nucleic acid amplification tests for Chlamydia trachomatis infection in women performed with an improved reference standard. J. Clin. Microbiol. 40:3757-3763.
Cuzick, J. 1985. A Wilcoxon-type test for trend. Stat. Med. 4:87-90.
Embil, J. A., H. J. Thiebaux, F. R. Manuel, L. H. Pereira, and S. W. MacDonald. 1983. Sequential cervical specimens and the isolation of Chlamydia trachomatis: factors affecting detection. Sex. Transm. Dis. 10:62-66.
Hadgu, A. 1998. A biomedical application of latent class models with random effects. Appl. Stat. 47:603-616.
Hernandez, T. J., K. L. Noller, and T. F. Smith. 1986. Detection of Chlamydia trachomatis using consecutive endocervical swabs. Prevalence in asymptomatic female adolescents and women attending a sexually transmitted disease clinic. J. Reprod. Med. 31:497-500.
Marrazzo, J. M., R. E. Johnson, T. A. Green, W. E. Stamm, J. Schachter, G. Bolan, E. W. Hook III, R. B. Jones, D. H. Martin, M. E. St. Louis, and C. M. Black. 2005. Impact of patient characteristics on performance of nucleic acid amplification tests and DNA probe for detection of Chlamydia trachomatis in women with genital infections. J. Clin. Microbiol. 43:577-584.
Munday, P. E., J. M. Carder, N. F. Hanna, and D. Taylor-Robinson. 1984. Is one swab enough to detect chlamydial infection of the cervix Br. J. Vener. Dis. 60:384-386.
Sternberg, M. R., and A. Hadgu. 2001. A GEE approach to estimating sensitivity and specificity and coverage properties of the confidence intervals. Stat. Med. 20:1529-1539.
Zeger, S. L., and K. Y. Liang. 1986. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 42:121-130.(Khalil G. Ghanem, Robert )