Background: High-risk human papillomavirus (HR-HPV) testing has become a preferred cervical cancer screening strategy in some countries due to its superior sensitivity over cytology-based methods for identifying cervical intraepithelial neoplasia of grade 2 or worse (CIN2+). Improved sensitivity has been accompanied by reductions in specificity and concerns regarding overscreening and overtreatment of women with transient or nonprogressing HR-HPV infections. Triage of HR-HPV+ women to colposcopy is, thus, warranted for appropriate management and treatment.

Methods: Using data from the Canadian Cervical Cancer Screening Trial (CCCaST), we compared the performance of cytology and HR-HPV strategies to detect CIN2+ among HR-HPV+ women (age, 30–69 years). Colposcopy referral rates and performance gains from adding other HR-HPV genotypes to HPV16/18+ triage were also evaluated.

Results: A strategy referring all women HPV16/18+ and HPV16/18, but with atypical squamous cells of undetermined significance or worse cytology (ASC-US+) had the highest sensitivity [82.5%; 95% confidence interval (CI), 70.9%–91.0%] but yielded the highest colposcopy referral rate. HPV16/18+ triage was the next most sensitive strategy (64.1%; 95% CI, 51.1%–75.7%). Low-grade squamous intraepithelial lesion or worse cytology (LSIL+) triage yielded a low sensitivity (32.8%; 95% CI, 21.9%–45.4%) but had the most favorable specificity (93.6%; 95% CI, 91.0%–95.6%), positive predictive value (41.5%; 95% CI, 28.1%–55.9%), and colposcopy referral rate of strategies examined. HPV viral load triage strategies did not perform optimally overall. Inclusion of HR-HPV genotypes 31 and 52 to HPV16/18+ triage provided the highest sensitivities.

Conclusion: Concerns surrounding HPV-based screening can be effectively mitigated via triage.

Impact: Balancing the benefits of HPV-based primary cervical screening with informed management recommendations for HR-HPV+ women may decide the success of its widening utilization. Cancer Epidemiol Biomarkers Prev; 26(6); 923–9. ©2017 AACR.

Virtually, all invasive cervical cancers are caused by persistent infection with certain genotypes of human papillomavirus (HPV), referred to as oncogenic or high-risk HPV (HR-HPV; ref. 1). Molecular techniques to detect the nucleic acid of these HR-HPV genotypes have demonstrated greater reliability and sensitivity for identifying cervical cancers and their high-grade precursors than cytology-based methods in cervical cancer screening (2, 3). As a result, cervical screening incorporating HR-HPV testing has become the preferred strategy in several high-income countries with organized (4) or opportunistic screening programs (5, 6).

Practical concerns regarding the adoption of HPV-based cervical screening remain, however. The improved sensitivity of HR-HPV testing comes with slight reductions in specificity (2), drawing concerns that such testing may lead to costly and unnecessary increases in follow-up and/or treatment of transiently HR-HPV+ women or those with nonprogressing HR-HPV infections. If founded, this would entail substantial burdens on a healthcare system, as well as the physical and psychological well-being of women concerned. Appropriate triage of HR-HPV+ women to colposcopy is, thus, necessary to identify those at immediate risk and prevent undue harm to those at low to moderate risk (i.e., the majority of HR-HPV+ women; ref. 7).

Using data from the Canadian Cervical Cancer Screening Trial (CCCaST; ref. 8, 9), we compared the performance of Papanicolaou (Pap) cytology and HR-HPV screening strategies to detect cervical intraepithelial neoplasia of grade 2 or worse histology (CIN2+) among HR-HPV+ women undergoing routine cervical screening in the greater Montreal area (Quebec) and St. John's (Newfoundland; Current Controlled Trials Number: ISRCTN57612064).

CCCaST design and participants

Women aged 30–69 years presenting for routine cervical cancer screening at any of the 30 participating clinics in the greater Montreal area and St. John's between 2002 and 2005 were invited to participate in CCCaST, a randomized controlled trial designed to compare HR-HPV DNA testing and Pap cytology as standalone screening strategies to detect CIN2+. Reports outlining the design of CCCaST and first screening round results have been published (8, 9); end-of-study and extended follow-up results are also available (10). Women were considered ineligible to participate if they met any of the following criteria: were under evaluation, treatment, or follow-up of a cervical lesion; were without a cervix; were pregnant; had a prior history of invasive cervical cancer; were unable to provide informed consent; or had received a Pap test within 12 months. A total of 10,154 women willing and eligible to enroll in the trial (4,400 in Montreal and 5,754 in St. John's) were randomized to 1 of the 2 screening strategies, depending on whether an HR-HPV test or a Pap test was guiding subsequent management and data analysis definitions (referred to as the index test). Both screening tests were included in each study arm due to ethical considerations at the time, but the sampling order was randomized to enable evaluations of index tests as if they had been done alone.

Screening test allocation was randomized 1:1 (stratified by clinic), and participants were blinded to this allocation. Cytotechnicians and cytopathologists examining Pap smears were unaware of CCCaST, enabling them to blindly and routinely evaluate samples. Colposcopists and pathologists ascertaining disease endpoints were blinded to screening test results, as were HPV testing laboratories. CCCaST underwent ethical review at McGill (Montreal) and Memorial (St. John's) Universities and at all participating hospitals and clinics. Sampling order of the screening tests was not shown to affect their performance during the first screening round of CCCaST (9); as such, for current analyses, study arms were collapsed to permit a cohort investigation in which all women received both screening tests at enrollment.

Screening tests

Pap smears were carried out through conventional means, with results reported or reclassified according to 2001 Bethesda System nomenclature (8, 9, 11). Atypical squamous cells of undetermined significance or worse cytology (ASC-US+) was considered abnormal. The Hybrid Capture 2 (HC2) Assay (Qiagen) was used to test cervical samples placed in specimen transport medium for HR-HPV status, and specimens were considered HR-HPV+ if the ratio of relative light units (RLU) of the sample to the mean RLUs of triplicates of positive control was ≥1 (equivalent to 1 pg of HPV DNA per milliliter). Neutralized HC2+ samples were further processed with a MasterPure DNA Purification Kit (Epicentre Biotechnologies), and extracted DNA was tested using the Linear Array HPV Genotyping Assay (LA PCR; Roche Diagnostics) to gather genotype-specific information (12).

Diagnostic procedures and case definitions

Women with an abnormal Pap or HR-HPV+ test result at enrollment were referred for a colposcopic exam. These examinations took place at participating clinics according to a standardized protocol to reduce verification bias (8, 9). The protocol included: (i) exocervical biopsies of all abnormal-appearing cervical regions; (ii) at least 1 exocervical biopsy of normal appearing cervices; and (iii) endocervical curettage (ECC). The integration of all colposcopy results was done at the study coordinating center, and diagnostic excisional procedures were carried out in cases of significant discrepancies. If a histologic diagnosis of CIN2, CIN3, adenocarcinoma in situ (AIS), squamous cell carcinoma, or adenocarcinoma was given on the basis of any histologic specimen (the collective reference standard defining case status; henceforth abbreviated CIN2+), women were managed appropriately and withdrawn from the study. Otherwise, a colposcopic examination was repeated after 6 months. Participants with normal test results at enrollment who were not selected for colposcopy and those who had colposcopic examinations and did not meet the case definition were invited for a second round of screening 12 to 18 months after their initial visit or last colposcopy visit (respectively).

Triage strategies

We evaluated the following triage algorithms for HR-HPV+ participants: (i) cytology: ASC-US+ threshold; (ii) cytology: low-grade squamous intraepithelial lesion or worse (LSIL+) threshold; (iii) HPV genotyping: HPV16+ threshold; (iv) HPV genotyping: HPV16+ and/or HPV18+ threshold (abbreviated HPV16/18+); and (v) as in strategy (iv), but with cytology at an ASC-US+ threshold conditional to testing HPV16/18. We also evaluated hypothetical triage strategies on the basis of the presumed HR-HPV viral load for cervical screening samples by examining the RLU signal obtained via HC2. We selected the following RLU thresholds: (vi) ≥25; (vii) ≥60; (viii) ≥150; and (ix) ≥500, so that proportions of women considered triage-positive would roughly approximate those associated with the above cytology and/or HPV genotyping triage algorithms. Strategies (i), (ii), (vi), (vii), (viii), and (ix) ignored HPV genotyping results; strategies (iii), (iv), (vi), (vii), (viii), and (ix) ignored cytology results; and strategies (vi), (vii), (viii), and (ix) ignored both HPV genotyping and cytology results. Strategy (v) was intended to reproduce a recently approved primary HPV screening algorithm based on HR-HPV genotyping and conditional cytology (13, 14).

Statistical analyses

We calculated sensitivity, specificity, positive predictive value (PPV), and the complement of the negative predictive value (cNPV = 1 − NPV) parameters, along with their associated 95% confidence intervals (CI), for CIN2+ detection for the above triage algorithms among women with valid results for the respective screening tests. As a supplemental analysis, these performance parameters were also calculated using CIN3+ as the disease endpoint (Supplementary Table S1). The cNPV parameter provides the probability of detecting disease among those testing negative during triage (also referred to as the false omission rate). Proportions of women that would be referred for colposcopy under these various algorithms (i.e., considered triage-positive) were also evaluated. In addition, we examined the value of considering positivity for HR-HPV genotypes 31, 33, 45, 52, and 58 when individually added to the HPV16/18+ genotyping triage algorithm via a receiver operating characteristic (ROC) plot (sensitivity vs. 1 − specificity). These genotypes were selected because of their designations as conferring the greatest risk (after HPV16/18) for cervical cancer worldwide (15, 16). Statistical analyses were performed using Stata 14.1 software (StataCorp).

Among the 10,120 women with valid HC2 results at enrollment, 614 (6.1%) tested HR-HPV+ (Montreal: 337; St. John's: 277). The median age of these HR-HPV+ women was 38 years (overall and by study center), and overall median follow-up time was 15.4 months (Montreal: 18.5 months; St. John's: 13.1 months). Sixty-eight of these HR-HPV+ women met the case definition (median follow-up time among cases: 5.3 months).

Valid cytology and histology results were available for 550 of the HR-HPV+ women at enrollment (Montreal: 294; St. John's: 256). While the majority of these women had normal cytology (443; 80.5%), 54 (9.8%) had ASC-US, 24 (4.4%) had LSIL, and 29 (5.3%) had high-grade squamous intraepithelial lesion or worse cytology (HSIL+) (Table 1). Case status was ascertained in 67 (12.2%) of these women, 32 of which had ASC-US+ and 22 of which had LSIL+ cytology. Assuming an ASC-US+ cytology triage threshold, 107 (19.5%) women would have been considered triage-positive (Table 2). Alternatively, 53 (9.6%) women would have been referred under a LSIL+ cytology triage threshold.

Table 1.

Histologic outcomes according to enrollment screening test results among HR-HPV+ women

Histology results
Enrollment screening test resultsNormalCIN1CIN2CIN3/AISCancerTotal
Cytology 
 NILM 341 67 19 15 443 
 ASC-US 29 15 54 
 LSIL 14 24 
 HSIL+ 13 29 
HPV genotyping 
 HPV16+ 48 12 15 18 94 
 HPV16 260 70 11 16 360 
 HPV16/18+ 71 18 18 21 130 
 HPV16/18 237 64 13 324 
HPV genotyping: reflex cytology (ASC-US+ threshold) 
 HPV16/18+: NILM 56 12 86 
 HPV16/18+: ASC-US+ 14 12 42 
 HPV16/18: NILM 199 48 258 
 HPV16/18: ASC-US+ 34 14 59 
HPV viral load (via HC2) 
 ≥25 RLU 137 47 14 21 224 
 <25 RLU 260 47 14 14 335 
 ≥60 RLU 99 35 10 12 160 
 <60 RLU 298 59 18 23 399 
 ≥150 RLU 66 28 109 
 <150 RLU 331 66 23 29 450 
 ≥500 RLU 31 14 55 
 <500 RLU 366 80 26 30 504 
Histology results
Enrollment screening test resultsNormalCIN1CIN2CIN3/AISCancerTotal
Cytology 
 NILM 341 67 19 15 443 
 ASC-US 29 15 54 
 LSIL 14 24 
 HSIL+ 13 29 
HPV genotyping 
 HPV16+ 48 12 15 18 94 
 HPV16 260 70 11 16 360 
 HPV16/18+ 71 18 18 21 130 
 HPV16/18 237 64 13 324 
HPV genotyping: reflex cytology (ASC-US+ threshold) 
 HPV16/18+: NILM 56 12 86 
 HPV16/18+: ASC-US+ 14 12 42 
 HPV16/18: NILM 199 48 258 
 HPV16/18: ASC-US+ 34 14 59 
HPV viral load (via HC2) 
 ≥25 RLU 137 47 14 21 224 
 <25 RLU 260 47 14 14 335 
 ≥60 RLU 99 35 10 12 160 
 <60 RLU 298 59 18 23 399 
 ≥150 RLU 66 28 109 
 <150 RLU 331 66 23 29 450 
 ≥500 RLU 31 14 55 
 <500 RLU 366 80 26 30 504 

NOTE: HPV testing was done using the HC2 assay. Pap cytology was done through conventional means according to 2001 Bethesda System nomenclature. Genotyping of HC2+ cervical samples was conducted using the linear array PCR assay. HPV viral load was determined using the HC2 assay.

Abbreviation: NILM, negative for intraepithelial lesion or malignancy.

Table 2.

Performance of triage strategies for the detection of CIN2+ among HR-HPV+ women

Triage algorithmThreshold for referralNumber referred (%)Sensitivity % (95% CI)Specificity % (95% CI)PPV % (95% CI)cNPVa % (95% CI)
Cytology triage ASC-US+ 107 (19.5) 47.8 (35.4–60.3) 84.5 (80.9–87.6) 29.9 (21.4–39.5) 7.9 (5.6–10.8) 
 LSIL+ 53 (9.6) 32.8 (21.9–45.4) 93.6 (91.0–95.6) 41.5 (28.1–55.9) 9.1 (6.7–11.9) 
HPV genotyping triage HPV16+ 94 (20.7) 53.1 (40.2–65.7) 84.6 (80.7–88.1) 36.2 (26.5–46.7) 8.3 (5.7–11.7) 
 HPV16/18+ 130 (28.6) 64.1 (51.1–75.7) 77.2 (72.7–81.3) 31.5 (23.7–40.3) 7.1 (4.6–10.5) 
HPV genotyping triage with reflex cytology if HPV16/18 HPV16/18+ or HPV16/18- with ASC-US+ 189 (42.3) 82.5 (70.9–91.0) 64.3 (59.3–69.1) 27.5 (21.3–34.5) 4.3 (2.2–7.5) 
HPV viral load triage ≥25 RLU 224 (40.1) 58.8 (46.2–70.6) 62.5 (58.1–66.8) 17.9 (13.1–23.5) 8.4 (5.6–11.9) 
 ≥60 RLU 160 (28.6) 38.2 (26.7–50.8) 72.7 (68.5–76.6) 16.3 (10.9–22.9) 10.5 (7.7–14.0) 
 ≥150 RLU 109 (19.5) 22.1 (12.9–33.8) 80.9 (77.1–84.2) 13.8 (7.9–21.7) 11.8 (9.0–15.1) 
 ≥500 RLU 55 (9.8) 14.7 (7.3–25.4) 90.8 (87.9–93.2) 18.2 (9.1–30.9) 11.5 (8.9–14.6) 
Triage algorithmThreshold for referralNumber referred (%)Sensitivity % (95% CI)Specificity % (95% CI)PPV % (95% CI)cNPVa % (95% CI)
Cytology triage ASC-US+ 107 (19.5) 47.8 (35.4–60.3) 84.5 (80.9–87.6) 29.9 (21.4–39.5) 7.9 (5.6–10.8) 
 LSIL+ 53 (9.6) 32.8 (21.9–45.4) 93.6 (91.0–95.6) 41.5 (28.1–55.9) 9.1 (6.7–11.9) 
HPV genotyping triage HPV16+ 94 (20.7) 53.1 (40.2–65.7) 84.6 (80.7–88.1) 36.2 (26.5–46.7) 8.3 (5.7–11.7) 
 HPV16/18+ 130 (28.6) 64.1 (51.1–75.7) 77.2 (72.7–81.3) 31.5 (23.7–40.3) 7.1 (4.6–10.5) 
HPV genotyping triage with reflex cytology if HPV16/18 HPV16/18+ or HPV16/18- with ASC-US+ 189 (42.3) 82.5 (70.9–91.0) 64.3 (59.3–69.1) 27.5 (21.3–34.5) 4.3 (2.2–7.5) 
HPV viral load triage ≥25 RLU 224 (40.1) 58.8 (46.2–70.6) 62.5 (58.1–66.8) 17.9 (13.1–23.5) 8.4 (5.6–11.9) 
 ≥60 RLU 160 (28.6) 38.2 (26.7–50.8) 72.7 (68.5–76.6) 16.3 (10.9–22.9) 10.5 (7.7–14.0) 
 ≥150 RLU 109 (19.5) 22.1 (12.9–33.8) 80.9 (77.1–84.2) 13.8 (7.9–21.7) 11.8 (9.0–15.1) 
 ≥500 RLU 55 (9.8) 14.7 (7.3–25.4) 90.8 (87.9–93.2) 18.2 (9.1–30.9) 11.5 (8.9–14.6) 

NOTE: HPV testing was done using the HC2 assay. Pap cytology was done through conventional means according to 2001 Bethesda System nomenclature. Genotyping of HC2+ cervical samples was conducted using the linear array PCR assay. HPV viral load was determined using the HC2 assay.

aThe false omission rate (1 − NPV; equivalently, cNPV) refers to the probability of detecting disease among women that test negative during triage.

HPV genotype–specific and histology results were available for 454 of the HR-HPV+ women at enrollment (Montreal: 253; St. John's: 201). Ninety-four (20.7%) of these women would have been triage-positive under an HPV16+ genotyping triage algorithm, whereas 130 (28.6%) would have met this definition under an HPV16/18+ threshold (Table 2). There were 64 (14.1%) women who met the case definition: 34 under the HPV16+ genotyping threshold and 41 under the HPV16/18+ threshold (Table 1). Assuming an HPV16/18+ genotyping triage strategy with conditional (reflex) cytology among women testing HPV16/18 (ASC-US+ referral threshold; n = 447), 189 (42.3%) women would have been referred for colposcopy. Fifty-two (11.6%) of these women met the case definition under this conditional strategy.

Valid HPV viral load and histology results were available for 559 of the enrollment HR-HPV+ women (Montreal: 297; St. John's: 262). Sixty-eight (12.2%) women met the case definition: 40 under the ≥25 RLU threshold, 26 under the ≥60 RLU threshold, 15 under the ≥150 RLU threshold, and 10 under the ≥500 RLU threshold (Table 1). Under these respective RLU thresholds, 224 (40.1%), 160 (28.6%), 109 (19.5%), and 55 (9.8%) women would have been triage-positive. As previously stated, these proportions of triage-positive women roughly approximated those associated with cytology and/or HPV genotyping triage algorithms (Table 2).

HPV16/18+ genotyping triage with reflex cytology among HPV16/18 women was clearly the most sensitive algorithm (82.5%; 95% CI, 70.9%–91.0%), although it would have referred the largest proportion of women to colposcopy (42.3%; Table 2); its sensitivity was statistically distinguishable from nearly every other strategy. HPV16/18+ genotyping triage alone, the next most sensitive strategy at 64.1% (95% CI, 51.1%–75.7%), would have referred 28.6% of women and differed statistically from the following algorithms: LSIL+ cytology triage (32.8%; 95% CI, 21.9%–45.4%); and HPV viral load triage with thresholds of ≥60 (38.2%; 95% CI, 26.7%–50.8%), ≥150 (22.1%; 95% CI, 12.9%–33.8%), and ≥500 RLUs (14.7%; 95% CI, 7.3%–25.4%). HPV16+ triage had a sensitivity of 53.1% (95% CI, 40.2%–65.7%) and was only statistically distinguishable from the latter 2 viral load algorithms. While HPV16+ and ASC-US+ triage would have referred roughly the same proportion of women to colposcopy (20.7% vs. 19.5%, respectively), the latter had a lower sensitivity (47.8%; 95% CI, 35.4%–60.3%).

Cytology triage with a LSIL+ threshold suffered from a lower sensitivity but would have referred the smallest proportion of women to colposcopy (9.6%) and had the highest specificity (93.6%; 95% CI, 91.0%–95.6%) and PPV (41.5%; 95% CI, 28.1%–55.9%) of the triage algorithms evaluated. Its specificity was distinguishable from nearly every other strategy. Among the algorithms with the highest sensitivity to detect CIN2+ [i.e., HPV genotyping, secondary (conditional), and cytology triage algorithms], specificity ranged from 64.3% (95% CI, 59.3%–69.1%) for conditional triage to 84.6% (95% CI, 80.7%–88.1%) for HPV16+ genotyping triage. HPV viral load triage strategies generally performed the worst in terms of PPV, as performance ranged from 13.8% (95% CI, 7.9%–21.7%) for ≥150 RLU to 18.2% (95% CI, 9.1%–30.9%) for ≥500 RLU thresholds. While viral load triage using a ≥25 RLU threshold yielded a higher sensitivity than many of the other strategies (58.8%; 95% CI, 46.2%–70.6%), it had the lowest specificity of all algorithms evaluated (62.5%; 95% CI, 58.1%–66.8%). Notably, HPV16/18+ genotyping triage with reflex cytology had the lowest cNPV (4.3%; 95% CI, 2.2%–7.5%); HPV viral load triage strategies performed poorly, with the exception of the ≥25 RLU threshold.

Figure 1 explores the respective gains in sensitivity to detect CIN2+ when HR-HPV genotypes 31, 33, 45, 52, and 58 were individually added to the HPV16/18+ genotyping triage algorithm, relative to their false positive rates (FPR; equivalently, 1 − specificity). Performance parameters for ASC-US+ cytology triage and HPV16/18+ genotyping triage are also shown for comparison. Although the sensitivities of the different genotyping algorithms were statistically indistinguishable, they were noticeably higher than that of ASC-US+ cytology triage. The latter outperformed the other strategies in terms of specificity, however (84.5%; 95% CI, 80.9%–87.6%). HPV16/18/31+ and HPV16/18/52+ genotyping algorithms had the highest (and identical) sensitivity to detect CIN2+ (73.4%; 95% CI, 60.9%–83.7%), with very similar specificity (65.9%; 95% CI, 61.0%–70.6% vs. 66.7%; 95% CI, 61.8%–71.3%). Sensitivity of these strategies was, indeed, distinguishable from that of ASC-US+ cytology triage. The incremental value of adding HPV45 to the HPV16/18+ triage algorithm in terms of sensitivity was also noticeable (70.3%; 95% CI, 57.6%–81.1% vs. 64.1%; 95% CI, 51.1%–75.7%) but came at a cost in specificity (71.5%; 95% CI, 66.8%–76.0% vs. 77.2%; 95% CI, 72.7%–81.3%).

Figure 1.

Sensitivity versus FPR (1 − specificity) for triage strategies among HR-HPV+ women. Sensitivity (with 95% CIs) versus the FPR (1 − specificity) for detection of CIN2+ among HR-HPV+ women for various triage strategies is shown. Sensitivity and specificity parameters, along with their associated 95% CIs, for the triage algorithms are presented in Supplementary Table S2. HPV testing was done using the HC2 assay. Pap cytology was done through conventional means according to 2001 Bethesda System nomenclature. Genotyping of HC2+ cervical samples was conducted using the linear array PCR assay. The proportion of women that would be referred for colposcopy under cytology triage at an ASC-US+ threshold is 19.5%; proportions for genotyping triage follow: HPV16/18+ = 28.6%; HPV16/18/31+ = 39.6%; HPV16/18/33+ = 32.4%; HPV16/18/45+ = 34.4%; HPV16/18/52+ = 39.0%; HPV16/18/58+ = 31.7%.

Figure 1.

Sensitivity versus FPR (1 − specificity) for triage strategies among HR-HPV+ women. Sensitivity (with 95% CIs) versus the FPR (1 − specificity) for detection of CIN2+ among HR-HPV+ women for various triage strategies is shown. Sensitivity and specificity parameters, along with their associated 95% CIs, for the triage algorithms are presented in Supplementary Table S2. HPV testing was done using the HC2 assay. Pap cytology was done through conventional means according to 2001 Bethesda System nomenclature. Genotyping of HC2+ cervical samples was conducted using the linear array PCR assay. The proportion of women that would be referred for colposcopy under cytology triage at an ASC-US+ threshold is 19.5%; proportions for genotyping triage follow: HPV16/18+ = 28.6%; HPV16/18/31+ = 39.6%; HPV16/18/33+ = 32.4%; HPV16/18/45+ = 34.4%; HPV16/18/52+ = 39.0%; HPV16/18/58+ = 31.7%.

Close modal

Cervical cancer screening has undergone foundational changes in recent years, with the paradigm of prevention efforts gradually shifting from cytology to HR-HPV testing. Harnessing the sensitivity of molecular-based screening, however, involves concerns regarding overscreening and overtreatment of women with transient or nonprogressing HR-HPV infections. In this report, we examined the performance of various cytologic and HR-HPV screening strategies to detect CIN2+ among HR-HPV+ women in an effort to investigate if concerns surrounding HPV-based screening can be effectively mitigated via triage.

A triage strategy referring all HPV16/18+ and HPV16/18 but ASC-US+ women had the highest sensitivity of all studied algorithms but would have referred the greatest proportion of women (42.3%) to colposcopy. This strategy, most akin to that recently approved in the United States with use of the cobas® 4800 HPV Test (Roche Molecular Systems Inc.) for primary cervical screening in women aged ≥25 years (13, 14), had a sensitivity that was distinguishable from nearly all others reported. In addition, it boasted the lowest false omission rate (cNPV) of all algorithms evaluated. HPV16/18+ genotyping triage (i.e., no adjunctive cytology) was the next most sensitive strategy. Expectedly, superior sensitivity came at a cost in terms of specificity. LSIL+ cytology triage performed the best in terms of the latter parameter, in addition to yielding the highest PPV and lowest colposcopy referral rate; however, its sensitivity was among the lowest. Similar findings for the sensitivity, specificity, and colposcopy referral rate of HPV16/18+ genotyping triage with adjunctive cytology were reported among HR-HPV+ women in a Dutch cohort study (VUSA-Screen; ref. 17), although PPV was higher in that study. Sensitivity and PPV of ASC-US+ cytology triage were also higher in that cohort, but colposcopy referral rates were comparable. Another Dutch study (POBASCAM; ref. 18) reported higher sensitivities and PPVs for these strategies than we observed. Both Dutch studies concluded that ASC-US+ cytology triage at baseline followed by repeat cytology within 12 months (possibly in combination with baseline HPV16/18+ genotyping in POBASCAM) was the most attractive strategy on account of its colposcopy referral rate, NPV, and PPV. A nested evaluation of a Swedish study (Swedescreen; ref. 19) indicated ASC-US+ cytology triage of HR-HPV+ women at baseline, with genotype-specific HR-HPV testing after 12 months for women with normal cytology, as the most effective strategy; notably, this study was limited to women aged 32–38 years.

Apart from an HPV viral load triage algorithm using a ≥25 RLU threshold, viral load triage strategies did not perform favorably—particularly with regard to sensitivity, PPV, and cNPV. Indeed, HR-HPV viral load has been considered as a potential biomarker that could improve the discriminatory ability of HPV-based testing to identify clinically meaningful rather than transient HR-HPV infections (20, 21). A sizeable body of research has consistently demonstrated an association between high HR-HPV viral load in cervical scrapes and prevalent histopathologic changes among HR-HPV+ women (22). A unique association between HPV16 viral load and incident high-grade lesions has also been demonstrated (23). However, demonstrating a consistent association between viral load and the severity of underlying lesions has proven more complicated. Reasoning for this complexity includes variability and frequency in cervical samplings and methodologies used to measure viral load, as well as the presence and extent of concurrent HR-HPV infections and/or diffuse and productive HR-HPV infections causing lower-grade lesions in the surrounding mucosa of higher-grade lesions (22–25). Such inherent heterogeneity may have at least partially accounted for the poor performance we observed.

Although the value of genotyping for HPV16 and HPV18 is now established in practice guidelines (5, 13), there is uncertainty regarding the extent to which adding other HR-HPV genotypes in the triage of HR-HPV+ women could enhance lesion detection. Our analyses demonstrated that the respective inclusion of HR-HPV genotypes 31 and 52 to the HPV16/18+ genotyping triage threshold provided the highest sensitivities to detect CIN2+ and that both strategies were statistically distinguishable from cytology triage with an ASC-US+ threshold. The relative importance of these genotypes has been evidenced in recent studies evaluating the management of HR-HPV+ women (26, 27). Nonetheless, the clinical usefulness of adding non-16/18 HR-HPV genotypes to an HPV16/18+ genotyping triage threshold would arguably have to help guide clinical management over and beyond that achievable by the latter threshold alone (5, 28). On the basis of our findings showing a lack of precision to support this condition, clinical usefulness of these additional genotypes in the triage of HR-HPV+ women remains questionable.

Ultimately, any screening strategy triaging HR-HPV+ women will need to be deemed sufficiently sensitive to be beneficial, while also reducing the potential harms of unnecessary follow-up and treatment of women with low disease risk. Current cervical screening guidelines in the United States have delineated the number of colposcopies as a primary measure of these harms (5). In our study, HPV16/18+ genotyping triage with reflex cytology if HPV16/18 was clearly the most sensitive algorithm but would have referred the greatest proportion of women to colposcopy. Whether this gain in sensitivity would be deemed acceptable in relation to the additional proportion of women that would necessarily undergo colposcopy is not (in itself) a hard and fast science. Indeed, there are no universally accepted risk thresholds for these various management strategies; acceptance involves a complicated interplay of sociocultural, infrastructural, and economic factors (7). In the Addressing the Need for Advanced HPV Diagnostics (ATHENA) study, this strategy was thought to provide an acceptable trade-off between sensitivity and the number of colposcopies performed (14); the strategy was subsequently recommended in published interim clinical guidance in the United States (13). A corollary to this trade-off is minimizing the false omission rate (cNPV), as described by Arbyn and colleagues more recently (29). The inherent purpose of triage is to maximize the probability of detecting meaningful disease in those testing positive; however, aiming to minimize the probability of failing to detect disease in those testing negative should not be overlooked.

Some important considerations for our study should be acknowledged. Since CCCaST enrolled women attending routine cervical screening, it is possible that women in the study differed from those in the general population who elect not to participate in screening. Namely, we would expect overall disease risks among enrolled participants to be lower than those among unscreened women. Although cervical cytology and histology were processed and interpreted using prevailing, gold-standard methodologies, misclassification of assessments may exist due to the subjective nature of these procedures (30). Blinding of observers to initial screening test results, however, provides some assurance that any potential misclassification would be impartial to any particular group(s) of women. By the same token, this blinding may have contributed to an underestimation of the sensitivity of cytology in a real-life triage setting, as its performance is affected by observers' knowledge of HR-HPV status (31, 32).

Notably, only cervical samples testing HC2+ were genotyped; of the 614 HC2+ women at enrollment, 116 (18.9%) had no LA PCR result (to include low-risk HPV genotypes). Cervical samples were processed and tested using LA PCR years after undergoing HC2 testing, which may have affected sample quality; in turn, low viral load infections present in these stored samples may have gone undetected. Although it would have been interesting to examine the value of considering positivity for other LA PCR genotypes in Fig. 1, limited statistical precision forced us to focus our attention on genotypes conferring the greatest risk (after HPV16/18) for cervical cancer worldwide (15, 16). To this point, evaluation of the etiologic importance of specific HR-HPV genotypes is inherently dependent on the endpoint for which a study is powered. Like many studies (2), CCCaST was powered to detect CIN2+, which naturally includes a mixture of HR-HPV–associated lesions that may/may not progress to invasive disease. Genotype-specific etiologic fractions for invasive cervical cancer may not be directly equivalent to those for CIN2+, however. Although CIN2+ remains an actionable endpoint for treatment in clinical practice in Canada and elsewhere, we acknowledge that results of these genotype-specific analyses should be interpreted with this point in mind.

Cervical disease detected early on in the follow-up period included mostly prevalent lesions; our analyses measured CIN2+ detection without attempting to assign exact dates of disease occurrence to simulate circumstances of clinical practice. CCCaST's standardized colposcopy protocol likely allowed us to detect more CIN2+ and less cervical cancer than would otherwise be expected in routine practice. Finally, RLU signal measurements obtained for HR-HPV+ cervical specimens provided only an aggregate estimate for viral load across the HR-HPV genotypes present in HC2, rather than more precise genotype-specific measures for viral load (22, 23). Given that these latter measures were not obtained, that our genotyping triage strategies inherently took account of the predictive values of individual HR-HPV genotypes, and that RLU measurements obtained via HC2 could easily be implemented into clinical practice, we elected to explore the value of RLU measurements as such.

CCCaST was conducted among women attending routine screening in routine practice, setting the stage for evaluations of screening effectiveness that necessarily reflect the complexities of a natural practice setting. Given cervical cancer etiology, it seems inevitable that utilization of HPV-based primary cervical screening will continue to widen. Appropriately balancing the benefits of this screening with informed management recommendations for HR-HPV+ women may decide the success of this paradigm shift.

A. Ferenczy is a consultant/advisory board member. S. Ratnam reports receiving commercial research grant from Roche Diagnostics and other commercial research support and speaker's bureau honoraria from Hologic Inc. E.L. Franco is a consultant/advisory board member for Roche and BD. M.-H. Mayrand is a Fonds de la Recherche du Québec - Santé (FRQS) Clinical Research Scholar. M.-H. Mayrand and her institution have received research funding related to her participation in a Merck HPV vaccine trial. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.D. Isidean, M.-H. Mayrand, A.V. Ramanakumar, I. Rodrigues, F. Coutlée, E.L. Franco

Development of methodology: S.D. Isidean, I. Rodrigues, F. Coutlée, E.L. Franco

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.D. Isidean, M.-H. Mayrand, I. Rodrigues, S. Ratnam, F. Coutlée, E.L. Franco

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.D. Isidean, A.V. Ramanakumar, F. Coutlée, E.L. Franco

Writing, review, and/or revision of the manuscript: S.D. Isidean, M.-H. Mayrand, I. Rodrigues, A. Ferenczy, S. Ratnam, F. Coutlée, E.L. Franco

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.D. Isidean, A. Ferenczy, E.L. Franco

Study supervision: S.D. Isidean, M.-H. Mayrand, E.L. Franco

Other (Biostatistics): A.V. Ramanakumar

This work was supported by the Canadian Institutes of Health Research (Team Grant CRN-83320 and MCT-54063 to E.L. Franco; Doctoral Award to S.D. Isidean). The Réseau SIDA et Maladies Infectieuses (SIDA-MI) du FRQS supported the optimization of HPV detection and genotyping.

*CCCaST Study Group: Montreal Research Staff: L. Abruzzese, K. Bellegarde, M. Bernardin, E. Duarte, F. Ferdinand, E.L. Franco (Principal Investigator), S.D. Isidean, G. Kelsal, M.H. Mayrand, M. Paquin, S. Piché, J. Poirier, A.V. Ramanakumar, A. Rodrigues, N. Rousseau, C. Schwartz, N. Slavtcheva, E. Tunistky. St. John's Research Staff: A. Batstone, G. Condon, A. Fitzpatrick, P. Francis, B. Halfyard, C. Head, C. Leonard, D. Mason, J. McGrath, V. Moulton, E. Oates, W. Shea. Montreal Clinical Collaborators: M.Y. Arsenault, G. Asselin, L. Authier, S. Bagga, P. Bastien, L. Bazinet, F. Beaudoin, P. Beaulieu, M.J. Bédard, S. Belinski, S. Bélisle, J. Benoit, M. Bernard, S. Bianki, L. Biron, F. Bissonnette, R. Bou-Habib, J. Bourque, B. Bradbury, M. Champagne, Y. Charles, P. Choquette, J.N. Couture, H.Q. Dao, C. Desjardins, J. Desjardins, L. Desrosiers, S. DiTommaso, L. Dontigny, M. Doyle, J. Dubé, M.J. Dupuis, F. Durocher, F. Engel, B. Fafard, A. Ferenczy, G. Fortier, A. Fortin, C. Fortin, D. Francoeur, D. Frechette, P. Fugère, G. Gagné, S. Gascon, M.J. Gaudreau, D. Gaudron, K. Gemayel, L. Gilbert, S. Gilbert, J. Gill, I. Girard, A. Gobeil, L. Granger, E. Grou, F. Grou, G. Guertin, J. Guimond, R. Hemmings, N. Ifergan, C. Johnson, L. Johnson, L. Ketchian, Y. Korcaz, C. Lafortune, J. Lalande, J.F. Lanctot, D. Landry, M. Landry, D. Langevin, I. Langlois, L. Lanmy-Monnot, L. Lapensée, L. Larouche, D. Laurin, M.C. Lavigne, Y. Lavoie, M. Leduc, F. Leger, N. Leroux, G. Luskey, N. Mansour, J. Marceau, A. Masse, I. Mayrand, M.H. Mayrand, L.R. McLauchlin, S. Menard, C. Mercer, M. Messier, B. Michon, M. Nadeau, M. Nguyen, S. Ouellet, C. Paquin, R. Paré, S. Peloquin, Y. Piché, R. Pichet, C. Rivard, I. Rodrigues, S. Roman, L. Rusimovic, G. Sanche, D. Soulière, D. Sproule, M. Steben, S. Still, D. Theriault, G. Tondreau, D. Tremblay, T. Minh Dung Vo, V.M. Whitehead, M. Yaffe, A. DiZazzo, C. Ziegler. St. John's Clinical Collaborators: E. Bannister, E. Callahan, J. Collingwood, P. Crocker, L. Dawson, A. Drover, J. Dunne, F. Fifield, J. Fitzgerald, D. Fontaine, B. Grandy, M. Greene, K. Halley, L. Hatcher, J. Hickey, P. Horwood, J. Janes, F. Jardine, L. Kieley, S. King, C. Kirby, N. Kum, E. Mate, S. McGrath, C. McManamon, K. Misik, M. O'Dea, P. O'Shea, C. Peddle, M. Penton, C. Pike, P. Power, L. Rogers, K. Saunders, P. Skirving, T. Sullivan, J. Verge, P. Wadden, M. Watson, M. Young. HPV DNA Testing Laboratories: F. Coutlée (Montreal), S. Ratnam (St. John's).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Walboomers
JM
,
Jacobs
MV
,
Manos
MM
,
Bosch
FX
,
Kummer
JA
,
Shah
KV
, et al
Human papillomavirus is a necessary cause of invasive cervical cancer worldwide
.
J Pathol
1999
;
189
:
12
9
.
2.
Arbyn
M
,
Ronco
G
,
Anttila
A
,
Meijer
CJ
,
Poljak
M
,
Ogilvie
G
, et al
Evidence regarding human papillomavirus testing in secondary prevention of cervical cancer
.
Vaccine
2012
;
30
:
F88
99
.
3.
Castle
PE
,
de Sanjose
S
,
Qiao
YL
,
Belinson
JL
,
Lazcano-Ponce
E
,
Kinney
W
. 
Introduction of human papillomavirus DNA screening in the world: 15 years of experience
.
Vaccine
2012
;
30
:
F117
22
.
4.
von Karsa
L
,
Arbyn
M
,
De Vuyst
H
,
Ronco
G
,
Schenck
U
,
Segnan
N
, et al
European guidelines for quality assurance in cervical cancer screening. Summary of the supplements on HPV screening and vaccination
.
Papillomavirus Res
2015
;
1
:
22
31
.
5.
Saslow
D
,
Solomon
D
,
Lawson
HW
,
Killackey
M
,
Kulasingam
SL
,
Cain
JM
, et al
American Cancer Society, American Society for Colposcopy and Cervical Pathology, and American Society for Clinical Pathology screening guidelines for the prevention and early detection of cervical cancer
.
Am J Clin Pathol
2012
;
137
:
516
42
.
6.
Murphy
J
,
Kennedy
EB
,
Dunn
S
,
McLachlin
CM
,
Fung Kee Fung
M
, et al
Cervical screening: a guideline for clinical practice in Ontario
.
J Obstet Gynaecol Can
2012
;
34
:
453
8
.
7.
Wentzensen
N
,
Schiffman
M
,
Palmer
T
,
Arbyn
M
. 
Triage of HPV positive women in cervical cancer screening
.
J Clin Virol
2016
;
76
:
S49
55
.
8.
Mayrand
MH
,
Duarte-Franco
E
,
Coutlee
F
,
Rodrigues
I
,
Walter
SD
,
Ratnam
S
, et al
Randomized controlled trial of human papillomavirus testing versus Pap cytology in the primary screening for cervical cancer precursors: design, methods and preliminary accrual results of the Canadian cervical cancer screening trial (CCCaST)
.
Int J Cancer
2006
;
119
:
615
23
.
9.
Mayrand
MH
,
Duarte-Franco
E
,
Rodrigues
I
,
Walter
SD
,
Hanley
J
,
Ferenczy
A
, et al
Human papillomavirus DNA versus Papanicolaou screening tests for cervical cancer
.
N Engl J Med
2007
;
357
:
1579
88
.
10.
Isidean
SD
,
Mayrand
MH
,
Ramanakumar
AV
,
Gilbert
L
,
Reid
SL
,
Rodrigues
I
, et al
Human papillomavirus testing versus cytology in primary cervical cancer screening: End-of-study and extended follow-up results from the Canadian cervical cancer screening trial
.
Int J Cancer
2016
;
139
:
2456
66
.
11.
Solomon
D
,
Davey
D
,
Kurman
R
,
Moriarty
A
,
O'Connor
D
,
Prey
M
, et al
The 2001 Bethesda System: terminology for reporting results of cervical cytology
.
JAMA
2002
;
287
:
2114
9
.
12.
Coutlee
F
,
Rouleau
D
,
Petignat
P
,
Ghattas
G
,
Kornegay
JR
,
Schlag
P
, et al
Enhanced detection and typing of human papillomavirus (HPV) DNA in anogenital samples with PGMY primers and the Linear array HPV genotyping test
.
J Clin Microbiol
2006
;
44
:
1998
2006
.
13.
Huh
WK
,
Ault
KA
,
Chelmow
D
,
Davey
DD
,
Goulart
RA
,
Garcia
FA
, et al
Use of primary high-risk human papillomavirus testing for cervical cancer screening: interim clinical guidance
.
Gynecol Oncol
2015
;
136
:
178
82
.
14.
Cox
JT
,
Castle
PE
,
Behrens
CM
,
Sharma
A
,
Wright
TC
 Jr
,
Cuzick
J
, et al
Comparison of cervical cancer screening strategies incorporating different combinations of cytology, HPV testing, and genotyping for HPV 16/18: results from the ATHENA HPV study
.
Am J Obstet Gynecol
2013
;
208
:
184 e1–184 e11
.
15.
de Sanjose
S
,
Quint
WG
,
Alemany
L
,
Geraets
DT
,
Klaustermeier
JE
,
Lloveras
B
, et al
Human papillomavirus genotype attribution in invasive cervical cancer: a retrospective cross-sectional worldwide study
.
Lancet Oncol
2010
;
11
:
1048
56
.
16.
Smith
JS
,
Lindsay
L
,
Hoots
B
,
Keys
J
,
Franceschi
S
,
Winer
R
, et al
Human papillomavirus type distribution in invasive cervical cancer and high-grade cervical lesions: a meta-analysis update
.
Int J Cancer
2007
;
121
:
621
32
.
17.
Rijkaart
DC
,
Berkhof
J
,
van Kemenade
FJ
,
Coupe
VM
,
Hesselink
AT
,
Rozendaal
L
, et al
Evaluation of 14 triage strategies for HPV DNA-positive women in population-based cervical screening
.
Int J Cancer
2012
;
130
:
602
10
.
18.
Dijkstra
MG
,
van Niekerk
D
,
Rijkaart
DC
,
van Kemenade
FJ
,
Heideman
DA
,
Snijders
PJ
, et al
Primary hrHPV DNA testing in cervical cancer screening: how to manage screen-positive women? A POBASCAM trial substudy
.
Cancer Epidemiol Biomarkers Prev
2014
;
23
:
55
63
.
19.
Naucler
P
,
Ryd
W
,
Tornberg
S
,
Strand
A
,
Wadell
G
,
Elfgren
K
, et al
Efficacy of HPV DNA testing with cytology triage and/or repeat HPV DNA testing in primary cervical cancer screening
.
J Natl Cancer Inst
2009
;
101
:
88
99
.
20.
Mesher
D
,
Szarewski
A
,
Cadman
L
,
Cubie
H
,
Kitchener
H
,
Luesley
D
, et al
Long-term follow-up of cervical disease in women screened by cytology and HPV testing: results from the HART study
.
Br J Cancer
2010
;
102
:
1405
10
.
21.
Kotaniemi-Talonen
L
,
Malila
N
,
Nieminen
P
,
Anttila
A
,
Tarkkanen
J
,
Laurila
P
, et al
Test positivity cutoff level of a high risk human papillomavirus test could be increased in routine cervical cancer screening
.
Int J Cancer
2008
;
123
:
2902
6
.
22.
Woodman
CB
,
Collins
SI
,
Young
LS
. 
The natural history of cervical HPV infection: unresolved issues
.
Nat Rev Cancer
2007
;
7
:
11
22
.
23.
Gravitt
PE
,
Kovacic
MB
,
Herrero
R
,
Schiffman
M
,
Bratti
C
,
Hildesheim
A
, et al
High load for most high risk human papillomavirus genotypes is associated with prevalent cervical cancer precursors but only HPV16 load predicts the development of incident disease
.
Int J Cancer
2007
;
121
:
2787
93
.
24.
Sherman
ME
,
Wang
SS
,
Wheeler
CM
,
Rich
L
,
Gravitt
PE
,
Tarone
R
, et al
Determinants of human papillomavirus load among women with histological cervical intraepithelial neoplasia 3: dominant impact of surrounding low-grade lesions
.
Cancer Epidemiol Biomarkers Prev
2003
;
12
:
1038
44
.
25.
Constandinou-Williams
C
,
Collins
SI
,
Roberts
S
,
Young
LS
,
Woodman
CB
,
Murray
PG
, et al
Is human papillomavirus viral load a clinically useful predictive marker? A longitudinal study
.
Cancer Epidemiol Biomarkers Prev
2010
;
19
:
832
7
.
26.
Cuzick
J
,
Ho
L
,
Terry
G
,
Kleeman
M
,
Giddings
M
,
Austin
J
, et al
Individual detection of 14 high risk human papilloma virus genotypes by the PapType test for the prediction of high grade cervical lesions
.
J Clin Virol
2014
;
60
:
44
9
.
27.
Schiffman
M
,
Burk
RD
,
Boyle
S
,
Raine-Bennett
T
,
Katki
HA
,
Gage
JC
, et al
A study of genotyping for management of human papillomavirus-positive, cytology-negative cervical screening results
.
J Clin Microbiol
2015
;
53
:
52
9
.
28.
Schiffman
M
,
Khan
MJ
,
Solomon
D
,
Herrero
R
,
Wacholder
S
,
Hildesheim
A
, et al
A study of the impact of adding HPV types to cervical cancer screening and triage tests
.
J Natl Cancer Inst
2005
;
97
:
147
50
.
29.
Arbyn
M
,
Roelens
J
,
Martin-Hirsch
P
,
Leeson
S
,
Wentzensen
N
, et al
Use of HC2 to triage women with borderline and mild dyskaryosis in the UK
.
Br J Cancer
2011
;
105
:
877
80
.
30.
Stoler
MH
,
Schiffman
M
,
Atypical Squamous Cells of Undetermined Significance-Low-grade Squamous Intraepithelial Lesion Triage Study G
. 
Interobserver reproducibility of cervical cytologic and histologic interpretations: realistic estimates from the ASCUS-LSIL Triage Study
.
JAMA
2001
;
285
:
1500
5
.
31.
Richardson
LA
,
El-Zein
M
,
Ramanakumar
AV
,
Ratnam
S
,
Sangwa-Lugoma
G
,
Longatto-Filho
A
, et al
HPV DNA testing with cytology triage in cervical cancer screening: Influence of revealing HPV infection status
.
Cancer Cytopathol
2015
;
123
:
745
54
.
32.
Bergeron
C
,
Giorgi-Rossi
P
,
Cas
F
,
Schiboni
ML
,
Ghiringhello
B
,
Dalla Palma
P
, et al
Informed cytology for triaging HPV-positive women: substudy nested in the NTCC randomized controlled trial
.
J Natl Cancer Inst
2015
;
107
.
pii: dju423
.