Screening trial of human papillomavirus for early detection of cervical cancer in Santiago, Chile

Authors

Errata

This article is corrected by:

  1. Errata: Erratum: Screening trial of human papillomavirus for early detection of cervical cancer in Santiago Volume 133, Issue 2, E1, Article first published online: 12 February 2013

Abstract

Cervical cancer mortality in Chile is four times higher than in developed countries. We compared the accuracy of human papillomavirus (HPV) DNA testing and conventional Papanicolaou (Pap) testing to detect prevalent precancerous and cancerous lesions in the routine clinical practice of the public health system. Women aged 25 years and older residing in the area covered by three primary care centers of Santiago, Chile, were invited to participate. Eligible women received both HPV DNA (Hybrid Capture 2) and Pap testing. Women positive by either test (Pap: ASCUS+, HC2: RLU/CO ≥1.0) underwent colposcopy and biopsy, as did a sample of double-negative women with an abnormal cervix at visual inspection or with risk factors for cervical lesions. Crude and verification bias-corrected sensitivities and specificities were estimated. In total, 8,265 women (98.8% of eligible) had complete screening results. Of these, 10.7% were HPV positive, 1.7% were Pap positive and 1.1% were positive by both tests. In all, 931 (11.3%) women were screen-positive, of whom 94.3% attended colposcopy. Additionally, 295 control women were invited for colposcopy, of whom 78% attended. In all, 42 CIN2, 45 CIN3 and 9 cancers were identified. Verification bias-corrected sensitivity for CIN2+ (95% confidence interval) was 92.7% (84.4–96.8) for HPV and 22.1% (16.4–29.2) for Pap; corresponding specificities were 92.0% (91.4–92.6) and 98.9% (98.7–99.0). In conclusion, in routine clinical practice in a developing country, HPV testing was four times more sensitive for CIN2+ than Pap testing, identifying three times more CIN2+ lesions; HPV testing was easily implemented in our established cervical cancer prevention program.

In Chile, cervical cancer causes the death of more than 600 women each year, mainly of low socio-economic level, and is the second most common cause of cancer death for Chilean women aged 20–44 years.1 The current country's mortality rate of 7.6/100,000 is four times higher than that of developed countries.2, 3 These facts emphasize the need to improve the effectiveness and equitability of the national cervical cancer prevention program.

There is ample evidence that the detection of human papillomavirus (HPV) DNA in cervical samples has a higher sensitivity for cervical cancer and precancerous lesions than the Papanicolaou (Pap) test,4 and high-quality HPV tests are routinely used in prevention programs in some developed countries. Although there have been many studies of implementing de novo screening methods in developing countries,5–8 there is less reported evidence of the effectiveness of incorporating HPV DNA detection into primary screening in an established cervical cancer prevention program in a middle-income developing country such as Chile. Furthermore, owing to the general belief that the Chilean Pap-based cervical cancer prevention program has been a success,9, 10 no efforts have been made to evaluate Pap testing accuracy or to explore alternative options to improve the program. This is the first study to evaluate the sensitivity and specificity of HPV testing and Pap testing in the routine clinical practice of public primary health centers in Chile.

Abbreviations:

ASCUS: atypical squamous cells of undetermined significance; CI: confidence interval; CIN: cervical intraepithelial neoplasia; CIN2+: CIN2 or worse; CIN3+: CIN3 or worse; HC2: Hybrid Capture 2; HPV: human papillomavirus; NPV: negative predictive value; Pap: Papanicolaou; PPV: positive predictive value; VIA: visual inspection with acetic acid

Material and Methods

Study setting and population

Three public health care centers in Santiago, Chile, and their referral hospital participated in the study, using the infrastructure, personnel and protocols already in place under the national cervical cancer prevention program. We recruited women aged 25–64 years residing locally, excluding women who were pregnant, hysterectomized or virgins. Women were invited to participate through an outreach campaign in the catchment area of each health center and, if interested, received an appointment with the health center midwife. Eligible women who agreed to participate signed an informed consent form and entered the study. Screening, diagnosis confirmation and treatment were free of charge for the participants.

Study procedures

Participants completed a brief survey on socio-demographics and cervical cancer risk factors, and underwent pelvic examination and the collection of samples for the screening tests: Pap cytology and HPV DNA testing; after the samples were obtained, the midwife performed a visual inspection of the cervix with acetic acid (VIA). All women with a positive HPV or Pap test, that is screen-positive women, were recalled for diagnosis confirmation via colposcopy and directed biopsy. Additionally, a control group of HPV- and Pap-negative women was also referred for colposcopy.

Specimen collection and cervical examination

All midwives were trained in the study protocol, after which sampling for both screening tests was added to their routine activities. A first sample was taken for cervical cytology, using an Ayre spatula plus cotton swab combination to obtain exfoliated ecto- and endocervical cells which were smeared and fixed onto a slide, and subsequently stained with standard Papanicolaou technique; samples were sent daily for analysis. A second sample was taken for HPV detection with the HC2 DNA Collection Device (Qiagen, Gaithersburg, MD), introducing the cytobrush into the cervical os, rotating it 360°, and then depositing it in a tube with transport medium; samples were stored in coolers and sent weekly for analysis.

Additionally, most midwives were trained to perform VIA in all women they screened. VIA was performed after the cervical sampling and consisted of the naked-eye visualization of the cervix after the application of a 3% acetic acid solution. A positive VIA was defined as the visualization of distinct, well-defined, dense acetowhite areas close to or abutting the squamocolumnar junction in the transformation zone or close to the external os if the squamocolumnar junction was not visible. VIA was considered undetermined when ill-defined, transparent, acetowhite areas were visualized in the transformation zone or close to the external os if the squamocolumnar junction was not visible.

Laboratory analysis

Pap smears were read by cytotechnicians and quality controlled by pathologists at the area referral hospital, using the Bethesda 2001 classification system; both were blind to the HPV test results. Samples with atypical cells of undetermined significance (ASCUS) or worse were considered positive. Samples with unsatisfactory results that were not repeated were considered negative.

HPV test samples were analyzed at the Molecular Biology Laboratory of the Pontificia Universidad Católica de Chile University by experienced personnel previously trained at the Qiagen facilities in Sao Paulo, Brazil; all were blind to the Pap test results. The test used was Hybrid Capture 2 (HC2) High-Risk HPV DNA test (Qiagen, Gaithersburg, MD), which is based on the detection of viral DNA by nucleic acid hybridization and is considered a standard test for HPV DNA screening.11 The technique identifies the presence of 13 oncogenic HPV genotypes (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 and 68). Samples were considered positive when the relative light unit/cutoff (RLU/CO) was ≥1.0, as recommended by the manufacturer; diagnostic accuracy at higher viral load cutoff points was analyzed.

Follow-up and diagnostic confirmation

Colposcopy and biopsy of any suspicious lesion was considered the gold standard for diagnostic confirmation. Women with a positive screening test, that is HPV or Pap, were referred for colposcopy as soon as the result was available. As previously mentioned, screen-negative women comprising the control group were also referred to colposcopy.

The reference center for colposcopies was the area base hospital, where gyneco-oncologists or gynecologists completing an oncology fellowship program performed the procedures as part of their regular activities. All physicians were blind to the screening status by study design; however, some participants may have informed them of their screening results during the visit. Histological examination of the biopsies was performed at the Pathological Anatomy Laboratory of the hospital. Pathologists were blind to the HPV test result, but not necessarily to the Pap test result, as the laboratory procedure is to check any previous Pap tests at the time of histological analysis. Detected lesions were treated according to the national cervical cancer prevention program. The diagnosis used for these analyses was the worst histological diagnosis on any biopsy, either at colposcopy or at treatment.

Statistical analysis

The total sample size was set to 8,597 women, assuming the following conditions based on Schatzkin et al12: a significance level of α = 0.05 for McNemar's test; sensitivities for the detection of CIN2+ lesions of 53% for the Pap test and 91.6% for the HPV test; a 90% power (1 − β) for detecting a 38.6% difference between tests; a 0.54% prevalence of CIN2+ lesions in the population; and a 10% loss to follow-up. The control group was a convenience sample set at 300 women.

We compared participants' sociodemographic characteristics by screening results and by lesion. Continuous variables are presented as mean and standard deviation and compared using the t-test. Categorical variables are presented as percentages and contrasted with the χ2 test.

Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) with 95% CI of both screening tests were calculated for CIN2+ and CIN3+ lesion detection. Crude estimates were obtained based on screen-positive women who attended colposcopy. To efficiently estimate the false-negative rate for CIN2+, given the small size of the control group, we selected the control women among those whose baseline risk of having a CIN2+ lesion was higher than that of the whole study population. The rationale for this is that the probability of both the HPV and the Pap tests missing a case in the whole population is very low, requiring a very large sample to identify these false negatives. However, by focusing the confirmatory test on women at higher risk, even a small sample might reveal enough missed cases. The high-risk control group was composed of women with abnormal VIA (all VIA positive women and a sample of VIA undetermined women) plus a random sample of women with normal VIA and with at least one risk factor for cervical lesions (age >35 years, >1 lifetime sexual partner, current smoker, ever use of oral contraceptives and unfaithful partner). Verification bias-adjusted estimates were calculated by using stratified sampling to weight up results from women who attended colposcopy, using the R Software CompareTests.13 The stratified sampling plan directly accounts for the fact that the screen-negative women referred to colposcopy were not a purely random sample, but were oversampled to have risk factors for cervical cancer, thereby adjusting the rate of missed cases to the actual risk profile of all the study participants. This approach to correcting verification bias is best applied when the selection of those on whom the confirmatory test is performed is under control of the researchers,14 as is our situation. CompareTests has previously been successfully applied to correct for verification bias in an analogous study comparing HPV, Pap and VIA tests in India.15

Results

From August 2009 to June 2010, a total of 8,407 women were invited to participate, of whom 0.5% were excluded because they were pregnant, hysterectomized or virgins. Of the eligible women, 98.8% completed the screening process and are the subjects of our study (Fig. 1).

Figure 1.

Enrollment, screening results and outcomes. *Includes 29 women with nonrepeated unsatisfactory Pap test results in whom two CIN2 and four CIN3 lesions were identified. **Includes 227 women with positive or undetermined VIA and 68 women with negative VIA but with demographic or behavioral risk factors for cervical cancer.

The mean age of the participants was 42.2 ± 10.3 years, with 97.6% between 25 and 64 years old; a small group of women were later identified as being outside of the age criteria; however, they were included in the study: 185 (2.2%) were aged 16–24 years and 14 (0.2%) were aged 65–78 years. The mean number of children per woman was 2.3 ± 1.3; 26.1% had less than nine years of schooling; 41.1% were smokers, with a mean of six cigarettes per day. Most women were married or cohabitating (69.9%), initiated sexual activity before 21 years old (75.0%), reported more than one lifetime sexual partner (63.9%), had never used condoms (78.8%), had ever used oral contraceptives (67.5%) and reported having had a Pap test in the past (93.3%). Women <30 years old initiated sexual activity at a younger age than women >49 years (mean ± SD, 18.1 ± 9.3 vs. 20.8 ± 11.1 years, p < 0.0001), and were more likely to have more than one lifetime sexual partner (72.8 vs. 60.8%, p < 0.0001) and to be current smokers (55.7 vs. 32.8%, p < 0.0001). When comparing women who tested positive by either screening test with those who tested negative by both tests, screen-positive women were significantly younger (38.7 ± 11.0 vs. 42.7 ± 10.1 years, p < 0.0001), initiated sexual activity at a younger age (18.9 ± 9.4 vs. 19.9 ± 10.3 years, p = 0.0024), and were more likely to be single or separated (47.3 vs. 27.9%, p < 0.0001), to have more than one lifetime sexual partner (81.3 vs. 61.6%, p < 0.0001), and to be current smokers (53.3 vs. 39.4%, p < 0.0001).

Figure 1 shows screening test results and outcomes. Of the 8,265 women studied simultaneously with HPV and Pap, 7,334 were negative by both screening tests. An 11.3% of all studied women had a positive HPV or Pap (screen-positive women). Positivity was 10.7% for HPV and 1.7% for Pap; 1.1% of women were positive by both tests. Among all women who were HPV positive, only 10.2% were also Pap positive; conversely, among those who were Pap positive, 65.2% were also HPV positive. Among the 931 screen-positive women, 94.3% attended colposcopy within one year of screening, of whom 35.9% required biopsy.

VIA was performed in 6,370 women (77.1% of participants) among whom 158 (2.5%) were positive, 260 (4.1%) were undetermined and 5,952 (93.4%) were negative. Among women who underwent VIA and were negative by both HPV and Pap (n = 5,672), 295 were referred for colposcopy as a control group as follows: all VIA positive women (n = 117), a sample of VIA undetermined women (n = 110, 47%), and a sample of VIA negative women with the previously mentioned risk factors (n = 68, 1.0%); 230 (78.0%) control women attended colposcopy.

Among screen-positive women who attended colposcopy, histology diagnosed 39 cases of moderate intraepithelial neoplasia (CIN2), 44 cases of severe intraepithelial neoplasia (CIN3) and eight cases of invasive cancer. Additionally, five CIN2+ lesions (three CIN2, one CIN3 and one cancer) were detected in the control women who attended colposcopy (all five were VIA positive). The HPV test detected 92.9% of CIN2 cases and 96.3% of CIN3+ cases, whereas the Pap test detected 28.6 and 38.9%, respectively. Of the 96 cases of CIN2+, the HPV test detected 94.8%, whereas the Pap test detected 34.3%; the HPV/Pap relative detection rate for CIN2+ was 2.8. Of the nine cancer cases, the HPV test detected eight, whereas the Pap test detected five; the only cancer missed by the HPV test was also missed by the Pap test, and was identified in the VIA positive subset of the control group.

Figure 2a shows the age distribution of screening test positivity. HPV positivity peaked at age <25 years, progressively decreasing to its lowest level by age 60, after which it had a small increase; Pap positivity also decreased after age 25. Figure 2b shows the age distribution of CIN3+ lesions detected by each screening test. Women <25 years old had the least CIN3+ lesions, although they had the highest screening positivity; conversely, women at age 45–49 had the highest lesion rate with the lowest screening positivity. HPV detected more lesions than Pap in all age groups.

Figure 2.

(a) Age distribution of test positivity among the 8,265 women screened with both HPV and Pap tests. (b) Age distribution of CIN3+ cases detected by each test among the 8,265 women screened with both HPV and Pap tests.

Table 1 summarizes diagnostic accuracy statistics for each screening test. Verification bias-corrected sensitivity and specificity for the detection of CIN2+ lesions were 92.7% (95% CI 84.4–96.8) and 92.0% (95% CI 91.4–92.6) for HPV, and 22.1% (95% CI 16.4–29.2) and 98.9% (95% CI 98.7–99.0) for Pap. Corrected NPVs for the detection of CIN2+ lesions were 99.8% (95% CI 99.5–99.9) for HPV and 97.8% (95% CI 97.1–98.3) for Pap. The uncorrected HPV/Pap relative PPV for CIN2+ was 0.44.

Table 1. Diagnostic accuracy of HPV1 and Pap testing for CIN2+ and CIN3+ lesions
inline image

When increasing the cutoff point of the HC2 HPV test to ≥2, sensitivity and specificity for the detection of CIN2+ lesions are 96.7% (95% CI 90.8–98.9) and 91.2% (95% CI 90.6–91.8), missing 3.1% of CIN2+ cases for an 8.7% decrease in the number of colposcopies required; when increasing the cutoff to ≥3, sensitivity and specificity for CIN2+ are 90.1% (95% CI 82.3–94.7) and 92.1% (95% CI 91.5–92.7), missing 9.4% of CIN2+ cases for a 17.3% decrease in colposcopy referrals.

Discussion

Our study demonstrated that under the current conditions of the public health service in Chile, HPV testing detects significantly more CIN2+ lesions than Pap testing (92.7 vs. 22.1%), at the cost of a higher false-positive rate (8.0 vs. 1.1%). HPV testing, as expected, had powerful sensitivity, but conventional cytology was unexpectedly insensitive, especially considering that cytology has been conducted by a long-established organized screening program. The Chilean National Cervical Cancer Prevention Program was introduced in 1987 as a pilot program in Santiago, the capital city of Chile, and became nationwide in 1994; it is based on a Pap test every three years for women aged 25–64 years who are beneficiaries of the public health system10 which corresponds to 70% of the female population in that age group; its coverage steadily increased from 50% in 1995 to 68% in 2004, after which it has remained relatively stable.16

As Pap test accuracy is highly dependent on training and expertise of the providers as well as workplace infrastructure, its sensitivity varies widely worldwide, with rates as low as 20% reported in Germany,17 26% in Peru,18 55% in Canada,19 57% in Africa and India,20 63% in Costa Rica,6 up to 77% in the United Kingdom,21 for the detection of CIN2+ lesions. In our study, Pap test sensitivity was lower than one would expect based on the downward trend of cervical cancer mortality in Chile (rates per 100,000 women 11.8 and 7.6 in 1990 and 2009),1 which has been attributed to the cervical cancer prevention program.

One explanation for this seemingly inconsistent result could be a decrease in Pap test quality owing to insufficient quality control. The study area is characterized by a rapid population growth, with a 48.6% increase between 1992 and 2002, whereas the overall country growth was only 15.2%,22 possibly causing an overload of the public health system and affecting test quality. There is no information available about Pap test performance in other areas of Chile as this is, to our knowledge, the first study of Pap sensitivity in the country.

On the other hand, it is possible that the frequent repetition of a poorly sensitive Pap test could lead to a decrease in mortality, given the slow progression of the disease,23 as lesions that are initially missed can be found at the next screening while they remain largely curable. The decrease in cervical cancer mortality observed in the last 20 years coincides with an improvement in the socioeconomic status of the Chilean population, with per capita income doubling in the period24 and female life expectancy at birth increasing from 76 years in 1990 to 81 years in 2007.25 It has been shown that in Latin America, per capita GDP is a greater predictor of cervical cancer mortality than cytology coverage.26 Nevertheless, Chile has a highly unequal income distribution, presenting one of the worst Gini coefficients of income distribution worldwide.27, 28 The high-income population is mainly served by a private health care system, whereas the rest is served by the Chilean public health system.29 Private sector practitioners do not always follow the Health Ministry guidelines and many perform yearly Pap tests. Accordingly, cervical cancer mortality has a highly unequal socioeconomic distribution in Chile, evidenced by a marked differential between women with less than four years of education and those with more than 12 years of education, with age-adjusted mortality rates per 100,000 women of 1.2 and 12.6, respectively, in 2009 (Fig. 3). Therefore, women from high-income groups, who undergo frequent screening, have cervical cancer mortality rates that are similar to those reported in developed countries, being a major contributor to the nationwide decreasing mortality curve, whereas low-income women still have mortality rates that are among the highest worldwide.2 In fact, cervical cancer mortality in Chile is the second worst of the OECD countries, although screening coverage is above the OECD average.30

Figure 3.

Cervical cancer mortality in Chilean women by level of education, 1990–2009. Age-adjusted rates per 100,000 women ≥15 years of age, using Chilean female population aged >15 years as standard (2002 National Census). ICD-10 codes: C53.0-C53.9. Source: Author's elaboration based on the Chile National Statistics Institute and the Chile Ministry of Health databases.

Although to a lesser extent, lower socio-economic levels have also experienced a decrease in cervical cancer mortality (Fig. 3). Another factor that could have contributed to this overall trend could be an increase in women's access to higher education, from 14% in 1990 to 41% in 2009, with the proportion of professional women in the workforce rising from 15 to 29% in the same period29 and a concurrent decrease in the fertility rate, from 2.4 to 1.9.31 There has also been an increase in access to quality healthcare, including the implementation of a new health reform plan in 2005 (Plan AUGE) which guarantees universal access to a standard of care for a number of diseases; this translates into shorter waiting times and better follow-up for women with suspected cervical cancer.

Our study confirms that Chilean women beneficiaries of the public health system (70% of the female population) are a high-risk population for cervical cancer, with a CIN2+ prevalence of 1.1% when screened with HPV testing, which is two to three times higher than the 0.4–0.5% prevalence reported in European trials similar to ours.32 Even when screened with Pap tests, the CIN2+ prevalence is higher in our study (0.4%) than in the same European studies (0.2–0.3%). Even with this higher background rate of lesions in Chile, Pap test positivity was very low in our study (1.7%) compared to the 3.4–6.8% positivity reported in those studies. This higher background CIN2+ prevalence in our setting may explain the higher positive predictive value of Pap in our study (24%) than that of European trials (4.0–9.8%).32 We should mention that a review of the databases of the Pap-based screening programs of several European countries33 shows that some countries have similar or higher CIN2+ rates than our study. Another explanation for the higher positive predictive value of Pap in our setting could be a lower proportion of equivocal or minor lesions among abnormal cytological results; when comparing our Pap test results with those of a large study of similarly aged women in the United States,34 we found comparable proportions of ASCUS (61.6 vs. 66.4%) and LSIL (25.4 vs. 29.2%), but a higher proportion of HSIL and cancer (13.0 vs. 4.4%); thus the ASCUS-HSIL/cancer ratio is lower in Chile than in developed countries (4.7 vs. 15.1), suggesting that Chilean pathologists avoid ASCUS diagnoses.

This trial used the infrastructure, personnel and protocols already in place under the cervical cancer prevention program, which suggests that our results reflect the real-world screening conditions in Chile. The existence for more than 20 years of an organized nationwide cervical cancer prevention program is a guarantee that an intervention to increase screening sensitivity, either by improving Pap test quality or by replacing it for a more sensitive test, would have an impact in the short term and with relatively little additional investment.

Often, the results of screening test studies conducted in developed countries are not replicated when studied in developing countries; this was the case of HC2 HPV testing, with reported sensitivities of 98.1% pooled estimate in developed countries, dropping to 50–88% in developing countries.4 Our results regarding HC2 sensitivity (92.7%) were comparable to those of developed countries. The main drawback of HC2 is the high rate of false positivity which would overload the reference level of the public health system causing excessive work-up, overdiagnosis and overtreatment. As we demonstrated, increasing the viral load cutoff for the HC2 HPV test is not an adequate strategy to reduce unnecessary referrals because of the important number of lesions that would be missed. Therefore, prior to introducing HPV-based primary screening, a triage method should be identified; alternatives include visual examination of the cervix, cytology and molecular biomarkers. Given the poor performance of the Pap test and VIA in our study, these would not be the first choice for triage in Chile unless their sensitivities were significantly improved. There are multiple molecular methods that are being researched throughout the world, including RNA-based tests, detection of HPV oncoproteins, expression of p16ink4a in cervical tissue and HPV16/18 genotyping. As the latter is currently available and affordable in Chile, we are at present studying the performance of HPV16/18 genotyping as a triage method in our population.

Women <30 years old had a high prevalence (1.0%) of CIN3+ lesions, which supports the need to screen women in this younger age group. This need is emphasized by the increasing lifetime number of sexual partners reported among Chilean women35 and by our finding that younger women had an earlier initiation of sexual activity, a greater number of sexual partners, and were more likely to be smokers; these behavioral changes could result in an increase in cervical cancer incidence, if the prevention program is not strengthened. However, the HPV false-positive rate for CIN3+ was much higher in women <30 years old than in women ≥30 years (19 and 9%), which suggests that younger women should be screened with a more specific test than HC2. As Pap test sensitivity in our study was very low (26.9% in women <30 years), cytology could be considered as a screening test only if its performance was improved to demonstrate a sensitivity of at least 70% as was reached in developed countries.36 The cost of the intervention needed to achieve this improvement in the performance of cytology should be balanced against the cost of introducing new screening tests.

Although the public health system has devoted considerable effort to improve Pap screening coverage, this has remained stable over the last years and has never reached the established goal of 80%. An HPV-based screening program might help improve coverage, as it would allow longer screening intervals that could increase compliance to screening recommendations and allow resources to be focused on fewer tests over the lifetime of women.

Our study was not designed to evaluate VIA for primary screening, but as a tool to select a high-risk control group. With this considered, and to avoid an excessive burden on the referral center, the midwives were instructed to use stringent criteria when classifying a VIA as abnormal, resulting in a low positivity rate (2.5%) compared to the studies aimed to evaluate the performance of VIA.20, 37 This may explain our low estimated sensitivity (35.7% for CIN2+ lesions), with an expected relatively high specificity (93.8%). Interesting to note is that, even with this stringent referral criterion, VIA identified five CIN2+ lesions that were missed by both HPV and Pap tests. Considering its ease of execution and low cost, VIA may be considered an alternative to cytology for primary screening or a triage method for HPV positive women, as has been discussed recently.38

One drawback of our study is that we did not perform the reference test in all screen-negative women, which would have been unfeasible, but in a subset of them. In most screening trials, the reference test has been applied to a random sample of screen-negative participants; however, in our study, we applied the reference test to a sample of women at higher risk for cervical lesions, that is those that are likely to have disagreeing test results. This novel strategy of selecting the sample to correct for verification bias has been demonstrated a highly efficient sampling design when the gold standard cannot be applied to all the screen-negative.13 Furthermore, the R package we used (CompareTests) is freely available and straightforward (http://dceg.cancer.gov/bb/tools/comparetests).

In conclusion, HPV testing showed a significantly higher sensitivity than Pap testing for the detection of cervical precancerous and cancerous lesions and it can be easily implemented into an established cervical cancer prevention program. However, further research into triage tests for HPV positive women is needed to address its lower specificity.

Acknowledgements

The HC2 DNA collection kits were donated by the manufacturing company Qiagen (Gaithersburg, MD); the analysis of the samples was funded by the grant. All other personnel and infrastructure related to screening, diagnosis and treatment were part of the regular coverage of the national healthcare system. The funding sources had no role in the study design, data collection, analysis and interpretation, or in the writing of this manuscript. The authors thank the following individuals for their assistance with our study: Mark Schiffman and Philip Castle of the National Cancer Institute, Bethesda, MD, USA, for their advice on study design. Martha Pruyas, Gloria Aguayo, Patricio Le Cerf and Eva Nielsen of the Department of Pathological Anatomy, Dr. Sotero Del Rio Hospital, Santiago, Chile for their support in the cytological and histological analyses. Ligia Valdivia and Eliana Romeo of the Department of Clinical Laboratories, Pontificia Universidad Católica de Chile, Santiago, Chile for their participation in the HPV analysis. Monica Alburquerque and Lorena Garay of the Department of Obstetrics and Gynecology, Dr. Sotero Del Rio Hospital, Santiago, Chile for their collaboration in the colposcopic examination. Javiera Leniz, Paz Cook, Patricia Tasama and Carolina Riveros of the Division of Public Health and Family Medicine, Pontificia Universidad Católica de Chile University, for their collaboration in the implementation of the trial. Pedro Vergara, Claudia Latorre, Mildred Rojas, Ximena Candia, Carlos Campos and Claudia Bejar of the Municipal Corporation of Puente Alto for facilitating the implementation of the study in their health facilities. The authors also thank the dedicated midwives who performed the screening examinations: Laura Leon, Claudia Duran, Gina Maza, Luz Medina and Anita Caceres of the Cardenal Raul Silva Henriquez Health Center; Karen Pavez, Maritza Benelli, Sonia Zuñiga, Karina Fuentes, Gilda Santis and Maritza Hernandez of the Padre Manuel Villaseca health center; Marcela Flores, Sonia Hernandez, Paula Reinoso, Andrea Cortes, Soledad San Martin, Veronica Vidal, Carolina Gatica, Pilar Monsalve, Pilar Matabenitez, Mirtha Rivera and Marilyn Caballero of the Alejandro del Rio health center, Puente Alto, Santiago, Chile.

Ancillary