- Email: [email protected]
Human papillomavirus detection for cervical cancer prevention with polymerase chain reaction in self-collected samples
Human papillomavirus detection for cervical cancer prevention with polymerase chain reaction in self-collected samples Felipe R. Lorenzato, MD,a,b Albert Singer, PhD, DPhil,c Linda Ho, MD, PhD,b Luiz Carlos Santos, MD,a Raimundo de Lucena Batista, MD,d Telma M. Lubambo, MD,a and George Terry, PhDb Recife and Pernambuco, Brazil, and London, United Kingdom OBJECTIVE: We studied the usefulness of self-sampling in cervical cancer prevention. STUDY DESIGN: A cross-sectional study was undertaken at screening services in Recife (Brazil); 253 women aged 16 to 88 years were included. Participants were randomly selected from a high-risk population for cervical neoplasia. All participants collected a self-sample with a cotton-tipped swab by rotating it against the vaginal epithelium and, possibly, the cervix. Physician-collected samples from the ectocervix and endocervix, respectively, with an Ayre’s spatula and a Cytobrush endocervical brush (Medscand) were followed by thorough colposcopy. Human papillomaviruses were detected by consensus polymerase chain reaction and typed by restriction fragment length polymorphism. RESULTS: The difference among human papillomavirus results in samples that were self-collected versus physician collected was significant (P < .03). The agreements were poor among patients with cervical intraepithelial neoplasia (CIN) grade 3 (κ <0.29) and cervical cancer (κ < 0.10). Self-sampling missed 50% more cancers than did physician sampling (P = .04). CONCLUSION: Self-sampling with a cotton-tipped swab for human papillomavirus detection is not a safe method for the collection of samples that are aimed at primary cervical cancer screening. (Am J Obstet Gynecol 2002;186:962-8.)
Key words: Self-sampling, HPV infection, cervical cancer prevention
The benefits of cervical cancer screening are essentially related to the proportion of the population that is covered and the efficacy of the method that is used. Satisfactory coverage has been achieved in developed countries, in which organized screening programs are well implemented. However, in developing countries, not only the inefficiencies in infrastructure but also the cultural taboos play a major role in the lack of effective measures to control and decrease cervical cancer incidence. The initial studies that involved animal models (rats) and humans that were undertaken by Stockard and Papanicolaou,1 Papanicolaou,2 and Papanicolaou and Traut3 were based on the analysis of cervical cells that were collected by cervicovaginal aspirates. Other studies From the Department of Obstetrics and Gynecology, Instituto Materno Infantil de Pernambucoa; the Department of Molecular Pathology, University College London,b the Department of Women’s Health, Whittington Hospital, University College London,c and the Pelvis Department, Hospital de Câncer de Pernambuco.d Supported by a grant (BEX 1178/96) from the Brazilian Government Teaching and Research Fund (CAPES). Received for publication May 21, 2001; revised September 5, 2001; accepted December 19, 2001. Reprint requests: Felipe Lorenzato, MD, Department of Molecular Pathology, University College London, 46 Cleveland St, London W1P 6DB, United Kingdom. E-mail: [email protected] © 2002, Mosby, Inc. All rights reserved. 0002-9378/2002 $35.00 + 0 6/1/122390 doi:10.1067/mob.2002.122390
962
about evaluations of cell collection for human papillomavirus (HPV) testing, including recent studies that involved pregnant women, have also used cervical vaginal lavage or a cotton-tipped or Dacron polyester swab as the method for collecting cervical cells.4-8 The rationale is that cells that are shed from the cervical epithelium travel through the vagina by gravity to reach the external environment. On the basis of the same principle, it is hypothesized that cells that are shed from the cervix, including cells from cervical intraepithelial neoplasia (CIN) and invasive cervical cancer, can be collected by women through self-collection with a cotton-tipped swab. This strategy, if proved effective, would be convenient to many women and could increase cervical cancer screening coverage in developing countries where taboos that are related to exposing the genitals to a male physician still lead to serious anxiety, which may result in noncompliance with organized cervical cancer screening programs. Can self-sampling for HPV detection and typing that is aimed at the improvement of cervical cancer prevention in a high-risk population provide enough cells from a representative proportion of the cervical ectocervix and canal to produce results that are matched in efficacy and safety to the cells that are collected by the current routine practice of cervical scrapes taken with an Ayre’s spatula and a Cytobrush endocervical brush (Medscand)? This study was designed to evaluate this query.
Lorenzato et al 963
Volume 186, Number 5 Am J Obstet Gynecol
Table I. Demographic characteristics of enrollees Variable Age (y)
Education
Age at 1st intercourse No. of sexual partners No. of pregnancies Parity
Stratification
No. of cases
Percentage
≤34 108 35-39 38 40-49 63 ≥50 44 None 47 Primary 71 Some high school 65 High school graduate 65 University graduate 4 < 16 years 67
42.9 15 24.9 17.4 18.6 28.4 25.7 25.7 1.6 26.5
1 2 3 ≥4 0 1-3 ≥4 0 1-3 ≥4
53.7 24.9 10.7 10.7 7.9 50.6 41.5 23.7 43.5 32.8
136 63 27 27 20 128 105 60 110 83
Table II. Comparison of self-sampling with physiciansampling results
Material and methods This was a cross-sectional study undertaken at 2 screening services in Recife (Brazil). The research protocol was reviewed and approved by the local Ethics Committee. The inclusion criterion was any woman who came for screening at the Institute of Mother and Child Health in Pernambuco and the Cancer Hospital in Pernambuco, including pregnant women, without age restriction, who were randomly selected to participate. The exclusion criteria were women without a cervix and women who were too ill or unwilling to participate in the study. After having the study details explained in lay terms, volunteers were invited to sign an informed consent form, at which time they received a sterile cotton-tipped swab to take a self-collected sample from the vagina and, possibly, cervix. The participants were advised to introduce the swab as far as it would go without leading to discomfort and to rub the swab around the vaginal epithelium in several rotatory movements. After feeling that most of the area had been covered, the participant should return the swab to a nurse who would disperse the cells into a universal tube that contained phosphate-buffered saline solution (PBS) and label the tube. The tubes were frozen at –20°C for later analysis at the Department of Molecular Pathology at University College London. The cotton-tipped swabs were different sizes, classified as small (8 cm), medium (15.2 cm) and large (21.5 cm, Figure). The different sizes of swabs were used because that was what was available at the time of the examination. A colposcopist collected samples for HPV testing with an Ayre’s spatula for the ectocervix and a Cytobrush endocervical brush for the endocervical canal just before performing a thorough colposcopic examination in each
HPV Positive Negative Total
Self-sampling (n)
Physician sampling (n)
Total (n)
58 195 253
73 180 253
131 375 506
participant, including women with a normal cervix, thus avoiding verification bias. The material that was collected was dispersed into a universal tube that contained phosphate-buffered saline solution and was frozen at –20°C. We used the L1 consensus pair of primers MY 09/11 for HPV DNA amplification and detection by polymerase chain reaction (PCR), typed by restriction fragment length polymorphism as previously described.9 The HPV types tested for were 6, 11, 42, 43, and 44 (categorized as low risk) and 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68 (categorized as high or intermediate risk), among others amplifiable by the MY09/11 pair of primers. The cytologic smears were processed by conventional cytologic examination and classified according to the revised Bethesda system. The histologic slides were processed according to standard guidelines and were read by local expert pathologists. A test to quantify the amount of DNA in each sample was not carried out in this study because the authors wanted to know the result as it would happen in routine clinical practice. Therefore, there was no insufficient HPV test result. None of the HPV DNA–positive samples had indeterminate types. The data were entered and analyzed by the SPSS 10 statistical package (SPSS, Inc, Chicago, Ill). The statistical methods used were mostly descriptive. The main results are presented by 2 2 contingency tables that compared the 2 methods of sample collection to the presence of HPV DNA. For the comparison of proportions, we used the Fisher exact test. The correlation of HPV detection by self-sampling and physician sampling was determined with an unweighted κ statistic to determine the percentage of correlation beyond that expected by chance. Statistical significance was considered when probability values were ≤.05. This study had a power of >80% within the 95% CI. Results A total of 253 women volunteered for the study. Five hundred six samples were analyzed; one half of the samples were self-collected by the participating women, and one half of the samples were collected by physicians. Among the samples that were collected by participating women, 200 samples (79%) were collected with a smallsized cotton-tipped swab; 10 samples (4%) were collected with a medium-sized cotton-tipped swab, and 43 samples (17%) were collected with a large-sized cotton-tipped
964 Lorenzato et al
May 2002 Am J Obstet Gynecol
Table III. HPV distribution by specific types HPV
Negative (n)
Self-collected Physician collected Total
195 180 375
6 (n)
11 (n)
16 (n)
18 (n)
11 5 16
4 2 6
21 33 54
1 2 3
Table IV. Agreements of results from self-sampling and physician sampling κ Results Normal Cervicitis CIN grade 1/condyloma CIN grade 2 CIN grade 3 SCC
Any HPV type 0.63 0.70 0.54 0.80 0.28 0.09
31 (n)
33 (n)
52 (n)
58 (n)
8 14 22
0 1 1
4 5 9
9 11 20
SCC, Squamous cell carcinoma.
swab. Participants’ ages ranged from 16 to 88 years, with a mean of 38.1 ± 13.7 years, a median of 38 years, and a mode of 28 years. Twenty-seven volunteers (10.7%) were pregnant at the time of enrollment. A better description of the demographic characteristics of the study population is found in Table I. Among the participants, 130 women (51.4%) each had a normal cervix; 44 women (17.4%) had cervicitis; 1 woman (0.4%) had condyloma; 12 women (4.7%) had CIN grade 1; 21 women (8.3%) had CIN grade 2; 28 women (11.1%) had CIN grade 3; and 17 women (6.7%) had cervical cancer. The overall HPV positivity of results from self-sampling and doctor-sampling were, respectively, 23% and 29%. There was no significant difference (P = .16) in the overall results from self-collected samples compared with physician-collected samples (Table II). However, as shown in Table III, the prevalence of high risk (HR) HPV among all self-collected samples was 17%, as opposed to 26% in samples collected by physicians (P < .03). Nevertheless, the concordance rate of positive and negative results between self-collected and physician-collected samples were, respectively, 80% and 92%. There were moderate agreements between the overall results from the samples that were self-collected and those that were physician collected with regard to positivity for HR HPV (κ 0.60) and positivity for any HPV type (κ 0.62). However, as shown in Table IV, our findings revealed poor agreements among patients with CIN grade 3 (κ <0.29) and extremely poor agreements among patients with invasive cervical cancer (κ <0.10). The size of the 3 different cotton-tipped swabs used for self-sampling (Figure) did not significantly influence the results of overall HPV detection (P = .33).
253 253 506
Table V. Number of CIN grade 1, 2, and 3 and invasive cancers detected by self-sampling and physician sampling
HR HPV 0.44 0.81 0.35 0.79 0.21 0.05
Total (n)
Self-sampling (n) Physician sampling (n)
CIN grade 1
CIN grade 2
5 5
12 14
CIN Cervical grade 3 cancer 12 22
7 14
One can even hypothesize that these 2 different methods of sample collection, self-sampling versus physician sampling, may actually be evaluating different topographic pathologic conditions, namely, vaginal versus cervical HPV infection. As a matter of fact, there was a significant statistical difference among the results of positive low risk and HR HPV infection in the samples that were analyzed from self-sampling versus physician sampling (P < .03). Table III shows the distribution of HPV types among samples that were self-collected and samples that were physician collected. When checking whether the age group may have influenced results, we observed that the discrepancies between self-sampling and physician sampling results occurred basically among women aged 35 to 50 years. In this age group, the prevalence of cervical neoplasia was higher, and the physician samplings picked up 30.8% more HR HPV infection than did self-samplings, although this difference was not significant (P = .42). This would, however, be the age range at which HPV positivity should be the lowest. Women are less likely to be HPV positive in the mid age range (highest when under the age of 30 years, with some upswing again after age 50 years). Even though most pregnant women would have not been in this age range, approximately 11% of participants were pregnant and therefore at higher risk of having HPV expression because of higher exposure to progesterone. Among women who used the small cotton-tipped swab for self-sampling, 16.5% tested positive for HR HPV. This figure was 40% and 20.9%, respectively, for the medium and large cotton-tipped swabs (P = .15). In the present study, HPV detection (as a means of identifying women who were at risk and therefore who should undergo colposcopic evaluation for possible biopsy and detecting cervical neoplasia) tested on samples self-collected resulted in 50% more missed diagnoses
Volume 186, Number 5 Am J Obstet Gynecol
of cervical cancer than did the analysis on samples that were collected with a spatula and Cytobrush (P = .04) and in 33.3% more missed diagnoses of CIN grade 2/3 (P = .56; Table V). Hence, a sample for cervical HPV detection that is collected with a spatula and a Cytobrush provides better results for primary cervical cancer screening than does self-sampling with a cotton-tipped swab. The prevalence of HPV infection in women with a normal cervix, based on the analysis of self-collected samples, was 8.9% compared with 12.8%, based on the analysis of physiciancollected samples (P > .71). Comment The use of a low-cost, yet accurate, method for the detection of cervical HPV infection is of utmost importance for improvements in cervical cancer screening programs based on cytologic examination. As seen in the present study, the use of self-sampling with simple cotton-tipped swabs after a clear explanation by a community health worker, particularly for the follow-up of women with latent HR HPV infection, could prove substantially valuable to overcome primary difficulties in the implementation of effective cervical cancer screening programs in developing countries. Some of the initial self-sampling studies for HPV detection used Southern blot or dot blot hybridizations, and the women collected the sample with a cervicovaginal lavage or vaginal tampons. An early study that included 48 women who self-collected samples using a cervicovaginal lavage compared the results of HPV detection with Southern blot hybridization to the results from physiciancollected samples using a wooden spatula (ectocervix) and a cotton-tipped swab (endocervix). The authors described a concordance of positive and negative results of 44% and 38%, respectively, and reported that the discordant results are unlikely to have occurred by chance (P = .004).10 The comparison of self-sampling with the use of a cervicovaginal lavage with physician sampling that also used a cervicovaginal lavage from 16 patients of the Bronx Municipal Hospital that was analyzed for the presence of HPV infection by Southern blot has shown concordance of positive and negative results of 75% and 83%, respectively, and no significant difference in overall results (P = 1.0).11 It has been reported that, among 48 women who self-collected a sample for HPV detection by inserting a tampon in the vagina and immediately withdrawing it compared with physician sampling with an Ayre’s spatula, there were correlations of 90% (κ 0.64) and 88% (κ 0.70) of positive results with dot blot and PCR, respectively, even though there was a significant difference in the overall results (P < .0002).12 Another early study that investigated the usefulness of self-sampling for HPV detection used the HPV profile method, which was an improvement of the ViraPap (Digene Diagnostic, Inc, Massachusetts) test. The author analyzed samples from 104 women and reported no signif-
Lorenzato et al 965
icant difference (P = .80) in the results of HPV positivity in samples self-collected with a standard 15.2-cm Dacron polyester swab compared with samples also collected with a Dacron polyester swab, but by nurse practitioners under direct observation of the cervix. The agreement was 91%.13 A study that involved 93 women aged 18 to 70 years that compared self-sampling by introducing and immediately removing a regular-sized tampon into the vagina with physician sampling of ectocervical and endocervical cells with a Dacron swab was undertaken in Philadelphia. Hybrid capture I (the tube test) was used for HPV detection. The authors reported a concordance of positive and negative results of 17.3% and 62.4%, respectively, a fair agreement (κ 0.49), and no significant difference in overall results (P > .1).14 A study that involved 247 women at high-risk for cervical disease and compared self-sampling with a Cytobrush that was introduced about 5 cm into the vagina for sample collection with physician sampling also with a Cytobrush, but under direct observation of the cervix, was undertaken in Germany.15 The authors used hybrid capture II for HPV detection and also showed a significant difference (P < .01) between the overall HPV positivity of results from self-sampling (53%) and physician sampling (42%). However, contrary to our findings, it was in selfcollected samples that the HPV detection tests resulted in a higher prevalence of HR HPV infection (47%), as opposed to in physician-collected samples (38%) through direct observation of the cervix (P < .05). A study undertaken in Cape Town, South Africa,16 that used a Dacron swab for self-sampling of black women between the ages of 35 and 65 years and a cone-shaped brush for a certified nurse to collect a sample from the cervix and a Dacron swab for a vaginal sample has been recently published. Hybrid capture II tests revealed 17.8% more positive results for the presence of HR HPV in cervical cells from samples collected with a brush compared with the HR HPV positivity in cells collected by selfsampling, which led to a significant difference (P < .05). They have also reported an 82% overall concordance of self-sampling and nurse-sampling results and stated that there was a moderate agreement between HPV results from self-collected vaginal cotton-tipped swabs and health personnel–collected Dacron-tipped vaginal swabs (κ 0.45). In another study, hybrid capture II analysis of samples that were collected from 200 women with a mean age of 31.5 years and that compared self-sampling with a Dacron swab from the vagina, vulva, and urine to physician sampling with a cone-shaped cervical brush from the transformation zone showed the following agreements: for urine (κ 0.41), vulval (κ 0.55), and vaginal samples (κ 0.76).17 The authors reported that the sensitivity of testing was progressively lower with increasing distance from the cervix and that the specificity rates for physician sam-
966 Lorenzato et al
pling from the cervix and self-sampling from the vagina were 52.1% and 53.5%, respectively. It was interesting to observe that, in the study by Wright et al,16 the HR HPV prevalence figures in women with a normal cervix that was analyzed by hybrid capture II were 17.1% and 15.5% for self-sampling and nursesampling results, respectively. In addition, among 280 women who had positive test results for HR HPV on selfsampling, 217 women (77.5%) had no cervical neoplasia; and among 275 women who tested positive for HR HPV on nurse sampling, 197 women (71.6%) also had no cervical neoplasia. These later figures in our study that used consensus PCR for HPV detection in high-risk women of any screening age were significantly lower: 8.3% (P = .0006) and 25% (P = .01), respectively. The former figures of HR HPV prevalence in women with a normal cervix described by Wright et al16 were much higher than the ones in our study, which were 1.6% (P = .007) and 5.9% (P < .08) for self-sampling and physician sampling, respectively. The prevalence rates of overall HPV types in our study were 8.9% and 12.8% for self-sampling and physician sampling results, respectively. As shown in Table I, the demographic characteristics of our study population did not differ substantially from the population in the mentioned study,16 except that we had no exclusion criteria regarding age for inclusion. In a recent study from Recife (Brazil) that involved women with a normal cervix who were aged 35 years or older, we observed a considerably elevated HR HPV positivity (58.6%) using a 1.0-pg cutoff value in samples that were collected with a Dacron swab by physicians under direct observation of the cervix and analyzed by hybrid capture I (the tube test).18 A study from Newfoundland (Canada) that used hybrid capture I for HPV detection in 69% of samples and hybrid capture II in the remaining 31% of samples also found low specificity (62%) for the identification of cervical neoplasias, more specifically, 51% for CIN grade 2/3.19 Our current findings are more in line with the 4.4% and 11% reported in a study that used general primers (GP) 5/6 PCR in samples from low-risk women aged 35 to 55 years and any screening age, respectively.20 The significantly lower and more realistic prevalence of HPV infection in women with a normal cervix, as assessed by PCR, is also corroborated by the 6% reported in another study that used GP 5+/6+ PCR, which included low-risk women aged 34 to 54 years.21 In other words, based on the facts, the difference between specificities of the 2 methods of HPV detection (hybrid capture II and PCR) is significant. These observations of satisfactory specificity for HPV detection that were achieved with PCR-based methods, in addition to the already established powerful PCR sensitivity, unveil important implications for improvements in cervical cancer screening programs and safe clinical guidelines. The use of HPV detection by PCR as a primary
May 2002 Am J Obstet Gynecol
Figure. The types of cotton-tipped swabs used for self-sampling.
screening tool for cervical cancer in combination with cytologic examination, as previously suggested,22 offers the highest standards of efficacy. So if there is a role for selfsampling in cervical cancer prevention, given the satisfactory specificity of PCR methods for HPV detection, women with latent cervical HR HPV infection (after thorough satisfactory colposcopy) should benefit from the convenience and safety of self-sampling as a practical method for yearly follow-up. This alternative follow-up strategy should provide reassurance for the women and their physicians of the persistence or regression of HPV infections. Follow-up studies that include some with selfsampling will help to set improved clinical guidelines and to establish the most appropriate routine screening interval (3, 5, or 10 years), with the use of HPV detection in samples that are collected by health professionals with a spatula and cytobrush as a primary screening tool. The difference between the results of HPV detection in samples that are self-collected and in samples that are collected by health personnel who directly observe the cervix, as found in the present study (P < .03), is corroborated by the findings of 4 previous studies.10,12,15,16 HR HPV positivity was higher in self-collected samples in 1 of the studies15; in another study,16 HR HPV positivity among women with a normal cervix was higher in self-collected samples than in health personnel–collected samples under direct observation of the cervix. Another investigation raises the issue that the method by which samples for HPV detection are collected may affect the results. An interesting study undertaken in Costa Rica that compared different variables that were associated with the results from HPV detection methods has shown that samples that were collected from the cervix with a conical brush provided a greater detection of HPV DNA than did the samples that were collected with a Dacron swab. However, with a sample size of 208 specimens, the difference was not statistically significant.23 Cotton-tipped swabs are perceived to be less successful in releasing trapped cells than Dacron swabs or brushes and to potentially interfere with molecular tests. This may partially explain the differences found in the present study. However, another investigation that also collected
Lorenzato et al 967
Volume 186, Number 5 Am J Obstet Gynecol
cervical cells with a cotton-tipped swab for HPV detection by PCR showed the presence of human β-globin sequences in 83% of collected specimens, thus suggesting that the samples were appropriately collected with cotton-tipped swabs.24 Cotton-tipped swabs that were used as collection devices have also been recently reported to be a satisfactory method of cell sampling for valid and useful PCR analyses in forensic medicine25,26 and other areas of applied health sciences.27-29 As described earlier, in some studies there may exist a moderate-to-good correlation between results from selfsampling and results from samples that are collected by doctors or health professionals in general. But, as shown in our study, there is a significant difference in the results of HPV detection that are analyzed by PCR when samples that were self-collected with a cotton-tipped swab are compared with samples that were collected by a physician who used a wooden spatula and a Cytobrush under direct observation of the cervix. This difference is supported by scientific evidence that the results of cervical HPV detection from samples that are collected with a brush are better than the samples collected with a Dacron swab.23 Moreover, the poor agreements that are related to CIN grade 3 and invasive cervical cancer (Table IV) prevent self-sampling from being used as a collection method for primary cervical cancer screening. These findings (Table IV) also suggest that CIN grade 3 and cervical cancer shed less cells that are infected with HPV DNA than do CIN grade 2, CIN grade 1, and subclinical HPV infection. CIN grade 2 was a real outlier compared with the other groups with true pathologic features. Although the κ values for CIN grades 1 and 3 were very poor, the κ values for CIN grade 2 were very high. The reason for this variance is unknown. Despite the potential benefits that self-sampling could bring to primary cervical cancer screening programs, the results of low risk and HR HPV detection by consensus PCR and typing by restriction fragment length polymorphism in self-collected samples were significantly different from the results in samples that were collected by physicians under direct observation of the cervix. Selfsampling with a cotton-tipped swab missed 50% more invasive cancers than did physician sampling with a spatula and an endocervical brush. The reliable and reassuring specificity that is provided by PCR methods in the analysis of self-collected samples offers a possible role for a selfcollection strategy to safely, conveniently, and effectively perform follow-up HPV detection in women with latent HR HPV infection. The data collection for this work was undertaken in Cervical Cancer Screening Programs at Instituto Materno Infantil de Pernambuco and Hospital do Câncer de Pernambuco, and the samples were analyzed at the Department of Molecular Pathology of the University College London.
REFERENCES
1. Stockard CR, Papanicolaou GN. A rhythmical “heat period” in guinea pigs. Science 1917;46:42-4. 2. Papanicolaou GN. New cancer diagnosis. Proceedings of the Third Race Betterment Conference; 1928. p. 528-34. 3. Papanicolaou GN, Traut HF. Diagnosis of uterine cancer by the vaginal smear. New York: Commonwealth Fund; 1943. 4. Morrison EAB, Gammon MD, Goldberg GL, Vermund SH, Burk RD. Pregnancy and cervical infection with human papillomaviruses. Int J Obstet Gynecol 1996;54:125-30. 5. Fife KH, Katz BP, Roush J, Handy VD, Brown DR, Hansell R. Cancer-associated human papillomavirus types are selectively increased in the cervix of women in the first trimester of pregnancy. Am J Obstet Gynecol 1996;174:1487-93. 6. Kemp EA, Hakenwerth AM, Laurent SL, Gravitt PE, Stoerker J. Human papillomavirus prevalence in pregnancy. Obstet Gynecol 1992;79:649-56. 7. Chang-Claude J, Schneider A, Smith E, Blettner M, Wahrensdorf J, Turek L. Longitudinal study of the effects of pregnancy and other factors on detection of HPV. Gynecol Oncol 1996;60: 355-62. 8. Tenti P, Zappatore R, Migliora P, Spinillo A, Maccarini U, Bendedittis M, et al. Latent human papillomavirus infection in pregnant women at term: a case-control study. J Infect Dis 1997;176:277-80. 9. Bernard HU, Chan SY, Manos MM, Ong CK, Villa LL, Delius H, et al. Identification and assessment of known and novel human papillomaviruses by polymerase chain reaction amplification, restriction fragment length polymorphism, nucleotide sequence and phylogenetic algorithms. J Infect Dis 1994;170:1077-85. 10. Vermund SH, Schiffman MH, Goldberg GL, Ritter DB, Weltman A, Burk RD. Molecular diagnosis of genital human papillomavirus infection: comparison of two methods used to collect exfoliated cervical cells. Am J Obstet Gynecol 1989;160:304-8. 11. Morrison EAB, Goldberg GL, Hagan RJ, Kadish AS, Burk RD. Self-administered home cervicovaginal: a novel tool for the clinical-epidemiologic investigation of genital human papillomavirus infections. Am J Obstet Gynecol 1992;167:104-7. 12. Fairley CK, Chen S, Tabrizi SN, Quinn MA, McNeil JJ, Garland SM. Tampons: a novel patient-administered method for the assessment of genital human papillomavirus infection. J Infect Dis 1992;165:1103-6. 13. Moscicki A-B. Comparison between methods for human papillomavirus DNA testing: a model for self-testing in young women. J Infect Dis 1993;167:723-5. 14. Harper DM, Hildesheim A, Cobb JL, Greenberg M, Vaught J, Lorincz AT. Collection devices for human papillomavirus. J Fam Pract 1999;48:531-5. 15. Hillemanns P, Kimmig R, Huttermann U, Dannecker C, Thaler CJ. Screening for cervical neoplasia by self-assessment for human papillomavirus DNA [letter]. Lancet 1999;354:1970. 16. Wright TC, Denny L, Kuhn L, Pollack A, Lorincz AT. HPV DNA testing of self-collected vaginal samples compared with cytologic screening to detect cervical cancer. JAMA 2000;283:81-6. 17. Sellors JW, Lorincz AT, Mahony JB, Mielzynska I, Lytwyn A, Roth P, et al. Comparison of self-collected vaginal, vulvar and urine samples with physician-collected cervical samples for human papillomavirus testing to detect high-grade squamous intraepithelial lesions. CMAJ 2000;163:513-8. 18. Lorenzato F, Singer A, Mould T, Santos LC, Maia A, Cariri L. Cervical cancer detection by hybrid capture and evaluation of local risk factors. Int J Gynecol Obstet 2001;73:37-42. 19. Ratnam S, Franco EL, Ferenczy A. Human papillomavirus testing for primary screening of cervical cancer precursors. Cancer Epidemiol Biomarkers Prev 2000;9:945-51. 20. Meijer CJLM, van den Brule AJC, Snijders PJF, Helmerhorst T, Kenemans P, Walboomers JMM. Detection of human papillomavirus in cervical scrapes by the polymerase chain reaction in relation to cytology: possible implications for cervical cancer screening. IARC Sci Publ 1992:119:271-82. 21. Rozendaal L, Walboomers JMM, van der Linden JC, Voorhorst FJ, Kenemans P, Helmerhorst TJM, et al. PCR-based high-risk HPV
968 Lorenzato et al
22.
23.
24.
25.
test in cervical cancer screening gives objective risk assessment of women with cytomorphologically normal cervical smears. Int J Cancer 1996;68:766-9. Lorenzato F, Ho L, Terry G, Singer A, Santos LC, Batista RL, et al. The use of human papillomavirus typing in detection of cervical neoplasia in Recife (Brazil). Int J Gynecol Cancer 2000;10:143-50. Peyton CL, Schiffman M, Lorincz AT, Hunt WC, Mielzynska I, Bratti C, et al. Comparison of PCR- and hybrid capture-based human papillomavirus detection systems using multiple cervical specimen collection strategies. J Clin Microbiol 1998;36:3248-54. Siegfried E, Rasnick-Conley J, Cook S, Leonardi C, Monteleone J. Human papillomavirus screening in pediatric victims of sexual abuse. Pediatrics 1998;1001:43-7. Della Manna A, Montpetit S. A novel approach to obtaining reliable PCR results from luminol treated bloodstains. J Forensic Sci 2000;45:886-90.
May 2002 Am J Obstet Gynecol
26. Hochmeister M, Rudin O, Meier R, Peccioli M, Borer U, Eisenberg A, et al. A foldable cardboard box for drying and storage of by cotton swab collected biological samples. Arch Kriminol 1997;3-4:113-20. 27. Zaal MJ, Volker-Dieben HJ, Wienesen M, D’Amaro J, Kijlstra A. Longitudinal analysis of varicella-zoster virus DNA on the ocular surface associated with herpes zoster ophthalmicus. Am J Ophthalmol 2001;131:25-9. 28. van Der Schee C, van Belkum A, Zwijgers L, van Der Brugge E, O’Neill EL, Luijendijk A, et al. Improved diagnosis of Trichomonas vaginalis infection by PCR using vaginal swabs and urine specimens compared to diagnosis by wet mount microscopy, culture, and fluorescent staining. J Clin Microbiol 1999;37:4127-30. 29. Buttner MP, Cruz-Perez P, Stetzenbach LD. Enhanced detection of surface-associated bacteria in indoor environments by quantitative PCR. Appl Environ Microbiol 2001;67:2564-70.
Key words: Self-sampling, HPV infection, cervical cancer prevention
The benefits of cervical cancer screening are essentially related to the proportion of the population that is covered and the efficacy of the method that is used. Satisfactory coverage has been achieved in developed countries, in which organized screening programs are well implemented. However, in developing countries, not only the inefficiencies in infrastructure but also the cultural taboos play a major role in the lack of effective measures to control and decrease cervical cancer incidence. The initial studies that involved animal models (rats) and humans that were undertaken by Stockard and Papanicolaou,1 Papanicolaou,2 and Papanicolaou and Traut3 were based on the analysis of cervical cells that were collected by cervicovaginal aspirates. Other studies From the Department of Obstetrics and Gynecology, Instituto Materno Infantil de Pernambucoa; the Department of Molecular Pathology, University College London,b the Department of Women’s Health, Whittington Hospital, University College London,c and the Pelvis Department, Hospital de Câncer de Pernambuco.d Supported by a grant (BEX 1178/96) from the Brazilian Government Teaching and Research Fund (CAPES). Received for publication May 21, 2001; revised September 5, 2001; accepted December 19, 2001. Reprint requests: Felipe Lorenzato, MD, Department of Molecular Pathology, University College London, 46 Cleveland St, London W1P 6DB, United Kingdom. E-mail: [email protected] © 2002, Mosby, Inc. All rights reserved. 0002-9378/2002 $35.00 + 0 6/1/122390 doi:10.1067/mob.2002.122390
962
about evaluations of cell collection for human papillomavirus (HPV) testing, including recent studies that involved pregnant women, have also used cervical vaginal lavage or a cotton-tipped or Dacron polyester swab as the method for collecting cervical cells.4-8 The rationale is that cells that are shed from the cervical epithelium travel through the vagina by gravity to reach the external environment. On the basis of the same principle, it is hypothesized that cells that are shed from the cervix, including cells from cervical intraepithelial neoplasia (CIN) and invasive cervical cancer, can be collected by women through self-collection with a cotton-tipped swab. This strategy, if proved effective, would be convenient to many women and could increase cervical cancer screening coverage in developing countries where taboos that are related to exposing the genitals to a male physician still lead to serious anxiety, which may result in noncompliance with organized cervical cancer screening programs. Can self-sampling for HPV detection and typing that is aimed at the improvement of cervical cancer prevention in a high-risk population provide enough cells from a representative proportion of the cervical ectocervix and canal to produce results that are matched in efficacy and safety to the cells that are collected by the current routine practice of cervical scrapes taken with an Ayre’s spatula and a Cytobrush endocervical brush (Medscand)? This study was designed to evaluate this query.
Lorenzato et al 963
Volume 186, Number 5 Am J Obstet Gynecol
Table I. Demographic characteristics of enrollees Variable Age (y)
Education
Age at 1st intercourse No. of sexual partners No. of pregnancies Parity
Stratification
No. of cases
Percentage
≤34 108 35-39 38 40-49 63 ≥50 44 None 47 Primary 71 Some high school 65 High school graduate 65 University graduate 4 < 16 years 67
42.9 15 24.9 17.4 18.6 28.4 25.7 25.7 1.6 26.5
1 2 3 ≥4 0 1-3 ≥4 0 1-3 ≥4
53.7 24.9 10.7 10.7 7.9 50.6 41.5 23.7 43.5 32.8
136 63 27 27 20 128 105 60 110 83
Table II. Comparison of self-sampling with physiciansampling results
Material and methods This was a cross-sectional study undertaken at 2 screening services in Recife (Brazil). The research protocol was reviewed and approved by the local Ethics Committee. The inclusion criterion was any woman who came for screening at the Institute of Mother and Child Health in Pernambuco and the Cancer Hospital in Pernambuco, including pregnant women, without age restriction, who were randomly selected to participate. The exclusion criteria were women without a cervix and women who were too ill or unwilling to participate in the study. After having the study details explained in lay terms, volunteers were invited to sign an informed consent form, at which time they received a sterile cotton-tipped swab to take a self-collected sample from the vagina and, possibly, cervix. The participants were advised to introduce the swab as far as it would go without leading to discomfort and to rub the swab around the vaginal epithelium in several rotatory movements. After feeling that most of the area had been covered, the participant should return the swab to a nurse who would disperse the cells into a universal tube that contained phosphate-buffered saline solution (PBS) and label the tube. The tubes were frozen at –20°C for later analysis at the Department of Molecular Pathology at University College London. The cotton-tipped swabs were different sizes, classified as small (8 cm), medium (15.2 cm) and large (21.5 cm, Figure). The different sizes of swabs were used because that was what was available at the time of the examination. A colposcopist collected samples for HPV testing with an Ayre’s spatula for the ectocervix and a Cytobrush endocervical brush for the endocervical canal just before performing a thorough colposcopic examination in each
HPV Positive Negative Total
Self-sampling (n)
Physician sampling (n)
Total (n)
58 195 253
73 180 253
131 375 506
participant, including women with a normal cervix, thus avoiding verification bias. The material that was collected was dispersed into a universal tube that contained phosphate-buffered saline solution and was frozen at –20°C. We used the L1 consensus pair of primers MY 09/11 for HPV DNA amplification and detection by polymerase chain reaction (PCR), typed by restriction fragment length polymorphism as previously described.9 The HPV types tested for were 6, 11, 42, 43, and 44 (categorized as low risk) and 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68 (categorized as high or intermediate risk), among others amplifiable by the MY09/11 pair of primers. The cytologic smears were processed by conventional cytologic examination and classified according to the revised Bethesda system. The histologic slides were processed according to standard guidelines and were read by local expert pathologists. A test to quantify the amount of DNA in each sample was not carried out in this study because the authors wanted to know the result as it would happen in routine clinical practice. Therefore, there was no insufficient HPV test result. None of the HPV DNA–positive samples had indeterminate types. The data were entered and analyzed by the SPSS 10 statistical package (SPSS, Inc, Chicago, Ill). The statistical methods used were mostly descriptive. The main results are presented by 2 2 contingency tables that compared the 2 methods of sample collection to the presence of HPV DNA. For the comparison of proportions, we used the Fisher exact test. The correlation of HPV detection by self-sampling and physician sampling was determined with an unweighted κ statistic to determine the percentage of correlation beyond that expected by chance. Statistical significance was considered when probability values were ≤.05. This study had a power of >80% within the 95% CI. Results A total of 253 women volunteered for the study. Five hundred six samples were analyzed; one half of the samples were self-collected by the participating women, and one half of the samples were collected by physicians. Among the samples that were collected by participating women, 200 samples (79%) were collected with a smallsized cotton-tipped swab; 10 samples (4%) were collected with a medium-sized cotton-tipped swab, and 43 samples (17%) were collected with a large-sized cotton-tipped
964 Lorenzato et al
May 2002 Am J Obstet Gynecol
Table III. HPV distribution by specific types HPV
Negative (n)
Self-collected Physician collected Total
195 180 375
6 (n)
11 (n)
16 (n)
18 (n)
11 5 16
4 2 6
21 33 54
1 2 3
Table IV. Agreements of results from self-sampling and physician sampling κ Results Normal Cervicitis CIN grade 1/condyloma CIN grade 2 CIN grade 3 SCC
Any HPV type 0.63 0.70 0.54 0.80 0.28 0.09
31 (n)
33 (n)
52 (n)
58 (n)
8 14 22
0 1 1
4 5 9
9 11 20
SCC, Squamous cell carcinoma.
swab. Participants’ ages ranged from 16 to 88 years, with a mean of 38.1 ± 13.7 years, a median of 38 years, and a mode of 28 years. Twenty-seven volunteers (10.7%) were pregnant at the time of enrollment. A better description of the demographic characteristics of the study population is found in Table I. Among the participants, 130 women (51.4%) each had a normal cervix; 44 women (17.4%) had cervicitis; 1 woman (0.4%) had condyloma; 12 women (4.7%) had CIN grade 1; 21 women (8.3%) had CIN grade 2; 28 women (11.1%) had CIN grade 3; and 17 women (6.7%) had cervical cancer. The overall HPV positivity of results from self-sampling and doctor-sampling were, respectively, 23% and 29%. There was no significant difference (P = .16) in the overall results from self-collected samples compared with physician-collected samples (Table II). However, as shown in Table III, the prevalence of high risk (HR) HPV among all self-collected samples was 17%, as opposed to 26% in samples collected by physicians (P < .03). Nevertheless, the concordance rate of positive and negative results between self-collected and physician-collected samples were, respectively, 80% and 92%. There were moderate agreements between the overall results from the samples that were self-collected and those that were physician collected with regard to positivity for HR HPV (κ 0.60) and positivity for any HPV type (κ 0.62). However, as shown in Table IV, our findings revealed poor agreements among patients with CIN grade 3 (κ <0.29) and extremely poor agreements among patients with invasive cervical cancer (κ <0.10). The size of the 3 different cotton-tipped swabs used for self-sampling (Figure) did not significantly influence the results of overall HPV detection (P = .33).
253 253 506
Table V. Number of CIN grade 1, 2, and 3 and invasive cancers detected by self-sampling and physician sampling
HR HPV 0.44 0.81 0.35 0.79 0.21 0.05
Total (n)
Self-sampling (n) Physician sampling (n)
CIN grade 1
CIN grade 2
5 5
12 14
CIN Cervical grade 3 cancer 12 22
7 14
One can even hypothesize that these 2 different methods of sample collection, self-sampling versus physician sampling, may actually be evaluating different topographic pathologic conditions, namely, vaginal versus cervical HPV infection. As a matter of fact, there was a significant statistical difference among the results of positive low risk and HR HPV infection in the samples that were analyzed from self-sampling versus physician sampling (P < .03). Table III shows the distribution of HPV types among samples that were self-collected and samples that were physician collected. When checking whether the age group may have influenced results, we observed that the discrepancies between self-sampling and physician sampling results occurred basically among women aged 35 to 50 years. In this age group, the prevalence of cervical neoplasia was higher, and the physician samplings picked up 30.8% more HR HPV infection than did self-samplings, although this difference was not significant (P = .42). This would, however, be the age range at which HPV positivity should be the lowest. Women are less likely to be HPV positive in the mid age range (highest when under the age of 30 years, with some upswing again after age 50 years). Even though most pregnant women would have not been in this age range, approximately 11% of participants were pregnant and therefore at higher risk of having HPV expression because of higher exposure to progesterone. Among women who used the small cotton-tipped swab for self-sampling, 16.5% tested positive for HR HPV. This figure was 40% and 20.9%, respectively, for the medium and large cotton-tipped swabs (P = .15). In the present study, HPV detection (as a means of identifying women who were at risk and therefore who should undergo colposcopic evaluation for possible biopsy and detecting cervical neoplasia) tested on samples self-collected resulted in 50% more missed diagnoses
Volume 186, Number 5 Am J Obstet Gynecol
of cervical cancer than did the analysis on samples that were collected with a spatula and Cytobrush (P = .04) and in 33.3% more missed diagnoses of CIN grade 2/3 (P = .56; Table V). Hence, a sample for cervical HPV detection that is collected with a spatula and a Cytobrush provides better results for primary cervical cancer screening than does self-sampling with a cotton-tipped swab. The prevalence of HPV infection in women with a normal cervix, based on the analysis of self-collected samples, was 8.9% compared with 12.8%, based on the analysis of physiciancollected samples (P > .71). Comment The use of a low-cost, yet accurate, method for the detection of cervical HPV infection is of utmost importance for improvements in cervical cancer screening programs based on cytologic examination. As seen in the present study, the use of self-sampling with simple cotton-tipped swabs after a clear explanation by a community health worker, particularly for the follow-up of women with latent HR HPV infection, could prove substantially valuable to overcome primary difficulties in the implementation of effective cervical cancer screening programs in developing countries. Some of the initial self-sampling studies for HPV detection used Southern blot or dot blot hybridizations, and the women collected the sample with a cervicovaginal lavage or vaginal tampons. An early study that included 48 women who self-collected samples using a cervicovaginal lavage compared the results of HPV detection with Southern blot hybridization to the results from physiciancollected samples using a wooden spatula (ectocervix) and a cotton-tipped swab (endocervix). The authors described a concordance of positive and negative results of 44% and 38%, respectively, and reported that the discordant results are unlikely to have occurred by chance (P = .004).10 The comparison of self-sampling with the use of a cervicovaginal lavage with physician sampling that also used a cervicovaginal lavage from 16 patients of the Bronx Municipal Hospital that was analyzed for the presence of HPV infection by Southern blot has shown concordance of positive and negative results of 75% and 83%, respectively, and no significant difference in overall results (P = 1.0).11 It has been reported that, among 48 women who self-collected a sample for HPV detection by inserting a tampon in the vagina and immediately withdrawing it compared with physician sampling with an Ayre’s spatula, there were correlations of 90% (κ 0.64) and 88% (κ 0.70) of positive results with dot blot and PCR, respectively, even though there was a significant difference in the overall results (P < .0002).12 Another early study that investigated the usefulness of self-sampling for HPV detection used the HPV profile method, which was an improvement of the ViraPap (Digene Diagnostic, Inc, Massachusetts) test. The author analyzed samples from 104 women and reported no signif-
Lorenzato et al 965
icant difference (P = .80) in the results of HPV positivity in samples self-collected with a standard 15.2-cm Dacron polyester swab compared with samples also collected with a Dacron polyester swab, but by nurse practitioners under direct observation of the cervix. The agreement was 91%.13 A study that involved 93 women aged 18 to 70 years that compared self-sampling by introducing and immediately removing a regular-sized tampon into the vagina with physician sampling of ectocervical and endocervical cells with a Dacron swab was undertaken in Philadelphia. Hybrid capture I (the tube test) was used for HPV detection. The authors reported a concordance of positive and negative results of 17.3% and 62.4%, respectively, a fair agreement (κ 0.49), and no significant difference in overall results (P > .1).14 A study that involved 247 women at high-risk for cervical disease and compared self-sampling with a Cytobrush that was introduced about 5 cm into the vagina for sample collection with physician sampling also with a Cytobrush, but under direct observation of the cervix, was undertaken in Germany.15 The authors used hybrid capture II for HPV detection and also showed a significant difference (P < .01) between the overall HPV positivity of results from self-sampling (53%) and physician sampling (42%). However, contrary to our findings, it was in selfcollected samples that the HPV detection tests resulted in a higher prevalence of HR HPV infection (47%), as opposed to in physician-collected samples (38%) through direct observation of the cervix (P < .05). A study undertaken in Cape Town, South Africa,16 that used a Dacron swab for self-sampling of black women between the ages of 35 and 65 years and a cone-shaped brush for a certified nurse to collect a sample from the cervix and a Dacron swab for a vaginal sample has been recently published. Hybrid capture II tests revealed 17.8% more positive results for the presence of HR HPV in cervical cells from samples collected with a brush compared with the HR HPV positivity in cells collected by selfsampling, which led to a significant difference (P < .05). They have also reported an 82% overall concordance of self-sampling and nurse-sampling results and stated that there was a moderate agreement between HPV results from self-collected vaginal cotton-tipped swabs and health personnel–collected Dacron-tipped vaginal swabs (κ 0.45). In another study, hybrid capture II analysis of samples that were collected from 200 women with a mean age of 31.5 years and that compared self-sampling with a Dacron swab from the vagina, vulva, and urine to physician sampling with a cone-shaped cervical brush from the transformation zone showed the following agreements: for urine (κ 0.41), vulval (κ 0.55), and vaginal samples (κ 0.76).17 The authors reported that the sensitivity of testing was progressively lower with increasing distance from the cervix and that the specificity rates for physician sam-
966 Lorenzato et al
pling from the cervix and self-sampling from the vagina were 52.1% and 53.5%, respectively. It was interesting to observe that, in the study by Wright et al,16 the HR HPV prevalence figures in women with a normal cervix that was analyzed by hybrid capture II were 17.1% and 15.5% for self-sampling and nursesampling results, respectively. In addition, among 280 women who had positive test results for HR HPV on selfsampling, 217 women (77.5%) had no cervical neoplasia; and among 275 women who tested positive for HR HPV on nurse sampling, 197 women (71.6%) also had no cervical neoplasia. These later figures in our study that used consensus PCR for HPV detection in high-risk women of any screening age were significantly lower: 8.3% (P = .0006) and 25% (P = .01), respectively. The former figures of HR HPV prevalence in women with a normal cervix described by Wright et al16 were much higher than the ones in our study, which were 1.6% (P = .007) and 5.9% (P < .08) for self-sampling and physician sampling, respectively. The prevalence rates of overall HPV types in our study were 8.9% and 12.8% for self-sampling and physician sampling results, respectively. As shown in Table I, the demographic characteristics of our study population did not differ substantially from the population in the mentioned study,16 except that we had no exclusion criteria regarding age for inclusion. In a recent study from Recife (Brazil) that involved women with a normal cervix who were aged 35 years or older, we observed a considerably elevated HR HPV positivity (58.6%) using a 1.0-pg cutoff value in samples that were collected with a Dacron swab by physicians under direct observation of the cervix and analyzed by hybrid capture I (the tube test).18 A study from Newfoundland (Canada) that used hybrid capture I for HPV detection in 69% of samples and hybrid capture II in the remaining 31% of samples also found low specificity (62%) for the identification of cervical neoplasias, more specifically, 51% for CIN grade 2/3.19 Our current findings are more in line with the 4.4% and 11% reported in a study that used general primers (GP) 5/6 PCR in samples from low-risk women aged 35 to 55 years and any screening age, respectively.20 The significantly lower and more realistic prevalence of HPV infection in women with a normal cervix, as assessed by PCR, is also corroborated by the 6% reported in another study that used GP 5+/6+ PCR, which included low-risk women aged 34 to 54 years.21 In other words, based on the facts, the difference between specificities of the 2 methods of HPV detection (hybrid capture II and PCR) is significant. These observations of satisfactory specificity for HPV detection that were achieved with PCR-based methods, in addition to the already established powerful PCR sensitivity, unveil important implications for improvements in cervical cancer screening programs and safe clinical guidelines. The use of HPV detection by PCR as a primary
May 2002 Am J Obstet Gynecol
Figure. The types of cotton-tipped swabs used for self-sampling.
screening tool for cervical cancer in combination with cytologic examination, as previously suggested,22 offers the highest standards of efficacy. So if there is a role for selfsampling in cervical cancer prevention, given the satisfactory specificity of PCR methods for HPV detection, women with latent cervical HR HPV infection (after thorough satisfactory colposcopy) should benefit from the convenience and safety of self-sampling as a practical method for yearly follow-up. This alternative follow-up strategy should provide reassurance for the women and their physicians of the persistence or regression of HPV infections. Follow-up studies that include some with selfsampling will help to set improved clinical guidelines and to establish the most appropriate routine screening interval (3, 5, or 10 years), with the use of HPV detection in samples that are collected by health professionals with a spatula and cytobrush as a primary screening tool. The difference between the results of HPV detection in samples that are self-collected and in samples that are collected by health personnel who directly observe the cervix, as found in the present study (P < .03), is corroborated by the findings of 4 previous studies.10,12,15,16 HR HPV positivity was higher in self-collected samples in 1 of the studies15; in another study,16 HR HPV positivity among women with a normal cervix was higher in self-collected samples than in health personnel–collected samples under direct observation of the cervix. Another investigation raises the issue that the method by which samples for HPV detection are collected may affect the results. An interesting study undertaken in Costa Rica that compared different variables that were associated with the results from HPV detection methods has shown that samples that were collected from the cervix with a conical brush provided a greater detection of HPV DNA than did the samples that were collected with a Dacron swab. However, with a sample size of 208 specimens, the difference was not statistically significant.23 Cotton-tipped swabs are perceived to be less successful in releasing trapped cells than Dacron swabs or brushes and to potentially interfere with molecular tests. This may partially explain the differences found in the present study. However, another investigation that also collected
Lorenzato et al 967
Volume 186, Number 5 Am J Obstet Gynecol
cervical cells with a cotton-tipped swab for HPV detection by PCR showed the presence of human β-globin sequences in 83% of collected specimens, thus suggesting that the samples were appropriately collected with cotton-tipped swabs.24 Cotton-tipped swabs that were used as collection devices have also been recently reported to be a satisfactory method of cell sampling for valid and useful PCR analyses in forensic medicine25,26 and other areas of applied health sciences.27-29 As described earlier, in some studies there may exist a moderate-to-good correlation between results from selfsampling and results from samples that are collected by doctors or health professionals in general. But, as shown in our study, there is a significant difference in the results of HPV detection that are analyzed by PCR when samples that were self-collected with a cotton-tipped swab are compared with samples that were collected by a physician who used a wooden spatula and a Cytobrush under direct observation of the cervix. This difference is supported by scientific evidence that the results of cervical HPV detection from samples that are collected with a brush are better than the samples collected with a Dacron swab.23 Moreover, the poor agreements that are related to CIN grade 3 and invasive cervical cancer (Table IV) prevent self-sampling from being used as a collection method for primary cervical cancer screening. These findings (Table IV) also suggest that CIN grade 3 and cervical cancer shed less cells that are infected with HPV DNA than do CIN grade 2, CIN grade 1, and subclinical HPV infection. CIN grade 2 was a real outlier compared with the other groups with true pathologic features. Although the κ values for CIN grades 1 and 3 were very poor, the κ values for CIN grade 2 were very high. The reason for this variance is unknown. Despite the potential benefits that self-sampling could bring to primary cervical cancer screening programs, the results of low risk and HR HPV detection by consensus PCR and typing by restriction fragment length polymorphism in self-collected samples were significantly different from the results in samples that were collected by physicians under direct observation of the cervix. Selfsampling with a cotton-tipped swab missed 50% more invasive cancers than did physician sampling with a spatula and an endocervical brush. The reliable and reassuring specificity that is provided by PCR methods in the analysis of self-collected samples offers a possible role for a selfcollection strategy to safely, conveniently, and effectively perform follow-up HPV detection in women with latent HR HPV infection. The data collection for this work was undertaken in Cervical Cancer Screening Programs at Instituto Materno Infantil de Pernambuco and Hospital do Câncer de Pernambuco, and the samples were analyzed at the Department of Molecular Pathology of the University College London.
REFERENCES
1. Stockard CR, Papanicolaou GN. A rhythmical “heat period” in guinea pigs. Science 1917;46:42-4. 2. Papanicolaou GN. New cancer diagnosis. Proceedings of the Third Race Betterment Conference; 1928. p. 528-34. 3. Papanicolaou GN, Traut HF. Diagnosis of uterine cancer by the vaginal smear. New York: Commonwealth Fund; 1943. 4. Morrison EAB, Gammon MD, Goldberg GL, Vermund SH, Burk RD. Pregnancy and cervical infection with human papillomaviruses. Int J Obstet Gynecol 1996;54:125-30. 5. Fife KH, Katz BP, Roush J, Handy VD, Brown DR, Hansell R. Cancer-associated human papillomavirus types are selectively increased in the cervix of women in the first trimester of pregnancy. Am J Obstet Gynecol 1996;174:1487-93. 6. Kemp EA, Hakenwerth AM, Laurent SL, Gravitt PE, Stoerker J. Human papillomavirus prevalence in pregnancy. Obstet Gynecol 1992;79:649-56. 7. Chang-Claude J, Schneider A, Smith E, Blettner M, Wahrensdorf J, Turek L. Longitudinal study of the effects of pregnancy and other factors on detection of HPV. Gynecol Oncol 1996;60: 355-62. 8. Tenti P, Zappatore R, Migliora P, Spinillo A, Maccarini U, Bendedittis M, et al. Latent human papillomavirus infection in pregnant women at term: a case-control study. J Infect Dis 1997;176:277-80. 9. Bernard HU, Chan SY, Manos MM, Ong CK, Villa LL, Delius H, et al. Identification and assessment of known and novel human papillomaviruses by polymerase chain reaction amplification, restriction fragment length polymorphism, nucleotide sequence and phylogenetic algorithms. J Infect Dis 1994;170:1077-85. 10. Vermund SH, Schiffman MH, Goldberg GL, Ritter DB, Weltman A, Burk RD. Molecular diagnosis of genital human papillomavirus infection: comparison of two methods used to collect exfoliated cervical cells. Am J Obstet Gynecol 1989;160:304-8. 11. Morrison EAB, Goldberg GL, Hagan RJ, Kadish AS, Burk RD. Self-administered home cervicovaginal: a novel tool for the clinical-epidemiologic investigation of genital human papillomavirus infections. Am J Obstet Gynecol 1992;167:104-7. 12. Fairley CK, Chen S, Tabrizi SN, Quinn MA, McNeil JJ, Garland SM. Tampons: a novel patient-administered method for the assessment of genital human papillomavirus infection. J Infect Dis 1992;165:1103-6. 13. Moscicki A-B. Comparison between methods for human papillomavirus DNA testing: a model for self-testing in young women. J Infect Dis 1993;167:723-5. 14. Harper DM, Hildesheim A, Cobb JL, Greenberg M, Vaught J, Lorincz AT. Collection devices for human papillomavirus. J Fam Pract 1999;48:531-5. 15. Hillemanns P, Kimmig R, Huttermann U, Dannecker C, Thaler CJ. Screening for cervical neoplasia by self-assessment for human papillomavirus DNA [letter]. Lancet 1999;354:1970. 16. Wright TC, Denny L, Kuhn L, Pollack A, Lorincz AT. HPV DNA testing of self-collected vaginal samples compared with cytologic screening to detect cervical cancer. JAMA 2000;283:81-6. 17. Sellors JW, Lorincz AT, Mahony JB, Mielzynska I, Lytwyn A, Roth P, et al. Comparison of self-collected vaginal, vulvar and urine samples with physician-collected cervical samples for human papillomavirus testing to detect high-grade squamous intraepithelial lesions. CMAJ 2000;163:513-8. 18. Lorenzato F, Singer A, Mould T, Santos LC, Maia A, Cariri L. Cervical cancer detection by hybrid capture and evaluation of local risk factors. Int J Gynecol Obstet 2001;73:37-42. 19. Ratnam S, Franco EL, Ferenczy A. Human papillomavirus testing for primary screening of cervical cancer precursors. Cancer Epidemiol Biomarkers Prev 2000;9:945-51. 20. Meijer CJLM, van den Brule AJC, Snijders PJF, Helmerhorst T, Kenemans P, Walboomers JMM. Detection of human papillomavirus in cervical scrapes by the polymerase chain reaction in relation to cytology: possible implications for cervical cancer screening. IARC Sci Publ 1992:119:271-82. 21. Rozendaal L, Walboomers JMM, van der Linden JC, Voorhorst FJ, Kenemans P, Helmerhorst TJM, et al. PCR-based high-risk HPV
968 Lorenzato et al
22.
23.
24.
25.
test in cervical cancer screening gives objective risk assessment of women with cytomorphologically normal cervical smears. Int J Cancer 1996;68:766-9. Lorenzato F, Ho L, Terry G, Singer A, Santos LC, Batista RL, et al. The use of human papillomavirus typing in detection of cervical neoplasia in Recife (Brazil). Int J Gynecol Cancer 2000;10:143-50. Peyton CL, Schiffman M, Lorincz AT, Hunt WC, Mielzynska I, Bratti C, et al. Comparison of PCR- and hybrid capture-based human papillomavirus detection systems using multiple cervical specimen collection strategies. J Clin Microbiol 1998;36:3248-54. Siegfried E, Rasnick-Conley J, Cook S, Leonardi C, Monteleone J. Human papillomavirus screening in pediatric victims of sexual abuse. Pediatrics 1998;1001:43-7. Della Manna A, Montpetit S. A novel approach to obtaining reliable PCR results from luminol treated bloodstains. J Forensic Sci 2000;45:886-90.
May 2002 Am J Obstet Gynecol
26. Hochmeister M, Rudin O, Meier R, Peccioli M, Borer U, Eisenberg A, et al. A foldable cardboard box for drying and storage of by cotton swab collected biological samples. Arch Kriminol 1997;3-4:113-20. 27. Zaal MJ, Volker-Dieben HJ, Wienesen M, D’Amaro J, Kijlstra A. Longitudinal analysis of varicella-zoster virus DNA on the ocular surface associated with herpes zoster ophthalmicus. Am J Ophthalmol 2001;131:25-9. 28. van Der Schee C, van Belkum A, Zwijgers L, van Der Brugge E, O’Neill EL, Luijendijk A, et al. Improved diagnosis of Trichomonas vaginalis infection by PCR using vaginal swabs and urine specimens compared to diagnosis by wet mount microscopy, culture, and fluorescent staining. J Clin Microbiol 1999;37:4127-30. 29. Buttner MP, Cruz-Perez P, Stetzenbach LD. Enhanced detection of surface-associated bacteria in indoor environments by quantitative PCR. Appl Environ Microbiol 2001;67:2564-70.