Early Detection and Diagnosis
High-risk human papillomavirus DNA load in a population-based cervical screening cohort in relation to the detection of high-grade cervical intraepithelial neoplasia and cervical cancer
Article first published online: 11 NOV 2008
Copyright © 2008 Wiley-Liss, Inc.
International Journal of Cancer
Volume 124, Issue 2, pages 381–386, 15 January 2009
How to Cite
Hesselink, A. T., Berkhof, J., Heideman, D. A.M., Bulkmans, N. W.J., van Tellingen, J. E.H., Meijer, C. J.L.M. and Snijders, P. J.F. (2009), High-risk human papillomavirus DNA load in a population-based cervical screening cohort in relation to the detection of high-grade cervical intraepithelial neoplasia and cervical cancer. Int. J. Cancer, 124: 381–386. doi: 10.1002/ijc.23940
- Issue published online: 14 NOV 2008
- Article first published online: 11 NOV 2008
- Manuscript Accepted: 7 AUG 2008
- Manuscript Received: 14 MAR 2008
- Netherlands Organization for Health and Development (ZonMW). Grant Number: 22000147
- viral load;
- real time PCR;
- cervical cancer
In a population-based cervical screening cohort, we determined the value of type-specific viral load assessment for the detection of high-grade cervical intraepithelial neoplasia and cervical cancer (≥CIN2). Viral load was determined by type-specific real-time PCR in women with single HPV16,-18,-31 and -33 infections, as determined by GP5+/6+-PCR. Study endpoints were the detection of cumulative ≥CIN2 or ≥CIN3 within 18 months of follow-up. High viral loads of HPV16,-31, and -33 were predictive for ≥CIN2 (relative risk of 1.6 (95% CI: 1.3–1.9), 1.7 (95% CI: 1.1–2.7) and 1.9 (95% CI: 1.1–3.1) per 10-fold change in viral load, respectively). For HPV18, the relative risk was of similar magnitude (1.5, 95% CI: 0.7–3.1), though not significant (p = 0.3). Subsequently, we determined the sensitivities of viral load for ≥CIN2 and ≥CIN3 in HPV DNA-positive women using viral load thresholds previously defined in a cross-sectional study. These thresholds were based on the 25th, 33rd and 50th percentiles of type-specific HPV16,-18,-31 or -33 viral load values found in women with normal cytology. For all types, combined sensitivities for ≥CIN2 were 93.5%, 88.8% and 77.7% for the 25th, 33rd and 50th percentile thresholds, respectively. Response-operator-characteristics (ROC) curve analysis showed that viral load testing on HPV DNA-positive women in addition to or instead of cytology may result in an increased sensitivity for ≥CIN2, but at the cost of a marked decrease in specificity in relation to cytology. Similar results were obtained when using ≥CIN3 as endpoint. In conclusion, in a cervical screening setting viral load assessment of HPV16, 18, 31 and 33 has no additive value to stratify high-risk HPV GP5+/6+-PCR-positive women for risk of ≥CIN2 or ≥CIN3. © 2008 Wiley-Liss, Inc.
A persistent infection with a high-risk human papillomavirus (hrHPV) is the key event in the development of cervical cancer and its precursor lesions, high-grade cervical intraepithelial neoplasia (≥CIN2).1–3 Several longitudinal population-based screening studies have shown that hrHPV testing, either or not in conjunction with cervical cytology, strongly enhances the sensitivity and thereby the negative predictive value for ≥CIN2 compared with cytology alone.4–9 As a consequence, the efficacy of cervical cancer screening can be improved by hrHPV testing, permitting longer screening intervals.9–11 The positive predictive value of hrHPV testing, however, needs improvement because a substantial number of hrHPV-positive women harbor transient infections that clear spontaneously and do not cause ≥CIN2.12 The inability of a hrHPV DNA test to discriminate between persistent and transient infections urges the need for additional biomarkers that enable the identification of hrHPV-positive women who are truly at risk of ≥CIN2. This may significantly reduce the number of women, particularly those with normal cytology, who are subjected to unnecessary clinical follow-up procedures based on a transient hrHPV infection.
In this context, several studies have investigated the value of the amount of viral DNA normalized per cervical scrape or per cell (i.e., viral load) as a risk determinant for high-grade CIN. For HPV16, it was found that increased viral loads are associated with an increased risk of ≥CIN213–16 and reduced amounts of viral DNA to the absence of ≥CIN2 and viral clearance.16, 17 Although for types other than HPV16 data were less conclusive,18–24 we previously managed in a cross-sectional study to define type-specific viral load thresholds for HPV16, -18, -31 and -33 above which all women with abnormal cytology and underlying ≥CIN3 lesions were discerned (i.e., a 100% sensitivity). These thresholds were based on the 33rd percentile of the lowest viral load values found in women with normal cytology participating in a population-based screening program.25 Based on these data it was suggested that viral load assessment allows excluding prevalent CIN3 lesions in 33% of hrHPV GP5+/6+-PCR positive women with normal cytology. However, it is not known whether the use of a viral load threshold on GP5+/6+-PCR-positive women has any impact on the sensitivity for high-grade CIN lesions and cervical cancer.
In this study, we evaluated in the population-based cervical screening trial POBASCAM4, 9, 26 viral load as a candidate biomarker to stratify GP5+/6+-PCR-based hrHPV-positive women for risk of high-grade CIN or worse diagnosed within 18 months of follow-up. Viral load assessment was determined by type-specific real-time PCR in all women who had single HPV16, -18, -31 or -33 infections at baseline. The sensitivities and specificities of viral load assessment for cumulative ≥CIN2 and ≥CIN3 diagnosed within 18 months of follow-up were determined using predefined type-specific viral load thresholds.25
Material and methods
Study population, specimen collection and HPV testing
The intake of the population-based randomized controlled cervical screening trial, i.e., POBASCAM,4 was between 1999 and 2002 and comprised 44,102 study participants. The study tested two screening policies, namely hrHPV testing by GP5+/6+-PCR-EIA combined with cytology (intervention group, n = 21,996) vs. cytology alone (control group, n = 22,106). Cytomorphological analysis was performed according to the CISOE-A classification, which can be translated easily to the Bethesda 2001 classification.27 Briefly, smears were classified as inadequate, normal, borderline dyskaryosis, mild dyskaryosis, moderate dyskaryosis, severe dyskaryosis, suspected of carcinoma in situ or suspected of invasive cancer. Borderline and mild dyskaryosis (BMD) equals ASC-US/ASC-H/LSIL and together moderate/severe dyskaryosis and suspected of carcinoma in situ (>BMD) equals high-grade squamous intraepithelial lesion.27, 28 In the control group, women were advised at the baseline round according to the current guidelines for cervical screening in The Netherlands.27 In the intervention group, repeat sampling at 6 and 18 months for cytology and HPV testing was advised to women with a hrHPV-positive normal smear and to women with BMD. Women with moderate dyskaryosis or worse (i.e., >BMD) were directly referred for colposcopy at any visit (i.e., baseline, 6 or 18 months) independent of hrHPV status. In case of hrHPV positivity at 18 months, women were referred for colposcopy-directed biopsy, irrespective of the cytology result. Women with normal cytology and a negative hrHPV test (hrHPV−) were recalled at the subsequent screening round (after 5 years). A more-detailed description of the study, the referral policy and follow-up procedure has been published previously.4, 9, 29
After a classic cervical smear was made on a glass slide, cervical scrapes were collected for hrHPV testing by placing the brush in 5 ml sterile phosphate-buffered saline and 0.05% merthiolate. Cervical samples, both baseline and, when applicable, follow-up smears at 6 and/or 18 months, were processed and subsequently tested for hrHPV by GP5+/6+-PCR-EIA.4 GP5+/6+-PCR-EIA-positive samples were subsequently typed by reverse line blot (RLB) according to previously described protocols.30
At baseline, a total of 1,102 women in the intervention group were hrHPV-positive by GP5+/6+-PCR-EIA of whom 472 women had a single infection with HPV16, -18, -31 and -33, as determined by RLB genotyping. Eligible for this study were women of whom sufficient material of the baseline cervical scrape was left for type-specific viral load assessment. Ultimately, 467 women were included comprising 256, 58, 109 and 44 women with single HPV16, -18, -31 and -33 infections at baseline, respectively. The numbers of women with single HPV16, -18, -31 or -33 infections stratified for cytology at baseline and histologic outcome are given in Table I. Of the 467 women included in the study, 91 women had >BMD and underwent colposcopy at baseline. Of the remaining 376 women, 163 had follow-up visits at both 6 and 18 months, 91 only at 6 months and 37 only at 18 months. The remaining 85 women had no follow-up. Written informed consents were obtained from all women. The study was approved by both the Medical Ethics Committee of the VU University Medical Center and the Ministry of Public Health.
Assessment of viral load
For viral load analysis, the high-pure PCR template preparation kit was used to extract DNA from 100 μl Tris-HCl sample, according to the recommendations of the manufacturer (Roche, Mannheim, Germany), except that samples were eluted in 100 μl elution buffer instead of 200 μl. Furthermore, DNA isolates were centrifuged twice for 1 min at 13,000 g and transferred to a new tube to ensure the complete removal of silica that may have co-eluted from the column.
HPV16, -18, -31 and -33 type-specific DNA load as well as the β-globin DNA load were determined by real-time PCR assays on the LightCycler instrument (Roche Diagnostics, Mannheim, Germany) in separate reactions to quantify the number of HPV copies and the number of cells, respectively. The LightCycler assays and the nucleotide sequences of the primers and probes are described in detail before.25, 31 The end volume of each real-time PCR reaction was 20 μl of which 5 μl (20 ng/μl) was extracted sample DNA. The LightCycler reactions were run in duplicate and values obtained were averaged. The viral load for each sample was calculated by dividing the HPV copy number by the number of cells and expressed as copies per cell (c/c).16
Study endpoints were the detection of cumulative ≥CIN2 or ≥CIN3 detected up to a follow-up period of 18 months. This consequently includes all ≥CIN2 or ≥CIN3 lesions detected at baseline as well. The relative risk of type-specific viral load for these endpoints was determined by Cox regression with correction for age. To enable calculations, GP5+/6+-PCR-positive samples that were negative in the LightCycler tests, thus containing numbers of HPV copies under the detection level, were set at <0.01 copies/cell.31, 32 HPV16, -18, -31 and -33 type-specific viral load values were log transformed yielding unskewed values. Women were censored when not attending repeat testing. Binary logistic regression analysis with correction for age, cytology, and HPV type was used to study the association between the baseline viral load, the change in viral load between baseline and moment of referral advice and study endpoints (i.e., ≥CIN2 or ≥CIN3). For further analyses, type-specific viral load values of HPV16, -18, -31 and -33 were dichotomized into “negative” and “positive” using one of three previously defined viral load thresholds, i.e., the 25th, 33rd and 50th percentile cut off point (Table II).25 These values were obtained using the lowest 25th, 33rd and 50th percentiles of the viral load values in women with normal cytology participating in population-based screening as cut off values. The overall data of viral load for a particular threshold (i.e., 25th, 33rd or 50th percentile) were obtained by summing up the dichotomized viral load data of each individual type. To compute the positive (PPV) and negative predictive values (NPV), we censored women when they did not attend repeat testing. The PPV and NPV were then computed using the Kaplan-Meier Risk estimate for right-censored data. Sensitivity and specificity estimates were computed by combining the PPV and NPV and the proportion of positive test scores.33 The sensitivity, specificity, PPV and NPV values were relative to the study population of HPV16, -18, -31 and -33 DNA-positive women. The median test was used to determine differences in median HPV load values in women with ≥CIN2 and ≥CIN3 between normal and abnormal cytology. The relation between sensitivity and specificity for ≥CIN2 or ≥CIN3 of the viral load thresholds and cytology was demonstrated in a response-operator-characteristics (ROC) curve. For this purpose, viral load values were also dichotomized for the 67th and 75th percentile threshold and cytology outcomes were dichotomized based on borderline, mild, or moderate dyskaryosis thresholds. The level of statistical significance was set at 0.05 and all analyses were performed using SPSS 11.5 software.
|Percentiles (c/c)1||HPV type|
Viral load is predictive for high-grade CIN or cervical cancer
The range of viral load values assessed in baseline cervical smears containing single HPV16, -18, -31 and -33 infections of women with and without high-grade CIN or worse diagnosed within 18 months of follow-up are depicted in Figure 1. The median viral loads in women with ≥CIN2 were 5.0 c/c (range <0.01–1.2 × 104) for HPV16, 5.0 c/c (range <0.01–55.6) for HPV18, 16.5 c/c (range 1.0–3.8 × 103) for HPV31, and 79.8 c/c (range: 3.6–3.6 × 103) for HPV33. The median viral load values of women without ≥CIN2 were lower for all 4 HPV types (i.e., 1.1 c/c (range: <0.01–456) for HPV16, 1.5 c/c (range: <0.01–202) for HPV18, 5.7 c/c (range <0.01–2.7 × 103) for HPV31, and 13.7 c/c (range: <0.01–241) for HPV33). In the following regression analysis age never showed a significant independent effect (p always ≥0.35). Nonetheless it was kept in the model to have age-adjusted relative risks (RR) for viral load. Among women with HPV16, -31 and -33 viral load values were predictive for ≥CIN2 (Cox regression; RR 1.6 (95% CI: 1.3–1.9), 1.7 (95% CI: 1.1–2.7), and 1.9 (95% CI: 1.1–3.1) per 10-fold change in viral load, respectively; Table III). For HPV18 the relative risk was of similar magnitude (1.5, 95% CI 0.7–3.1), although not significant (p = 0.3). After repeating the analyses with ≥CIN3 as endpoint similar results to those with ≥CIN2 were obtained (data not shown), although the association between viral load and ≥CIN3 was only significant for HPV16 (Cox regression; RR 1.6 (95% CI: 1.3–2.0) per 10-fold change in viral load). We also examined the effect of a change in viral load between baseline and the moment of referral advice on the risk of ≥CIN2. For types 16, 18, 31 and 33 combined, but not at single level, we found a marginal, positive association (logistic regression; RR 1.7 (95% CI: 1.0–3.0) per 10-fold change in viral load, p = 0.06). For ≥CIN3 this association was slightly lower (RR 1.4 (95% CI 0.9–2.1) per 10-fold change in viral load, p = 0.12).
|HPV type||≥CIN2/total number of infections||Cox regression|
|Relative risk1||95% CI|
Clinical performance of viral load thresholds for high-grade CIN or worse
The sensitivity and specificity for 18-month cumulative ≥CIN2 were determined after applying the previously defined viral load thresholds (Table II).25 To allow these analyses, the type-specific viral load values were dichotomized using the thresholds of either the 25th, 33rd, or 50th percentile. The overall status for the percentile thresholds was the combination of the dichotomous values of the 4 types together. Table IV shows for HPV16, -18, -31, -33 and the 4 HPV types together the sensitivity, specificity, PPV, and NPV estimates for the 33rd percentile threshold to detect ≥CIN2/3 lesions. For HPV16, these values for ≥CIN2 were 88.2% (95% CI: 80.5–93.1), 33.1% (95% CI: 29.8–36.5), 55.9% (95% CI 47.8–64.0), and 74.7% (95% CI 60.3–89.1), respectively. These estimates differed only slightly for the other 3 HPV types separately or after pooling data from all types. At the 25th and 50th percentile viral load thresholds sensitivity, specificity, PPV, and NPV estimates for ≥CIN2 also displayed a similar trend among the 4 HPV types. The pooled, overall values for ≥CIN2 were for the 25th and 50th percentiles: sensitivity: 93.5% (95% CI: 88.3–96.5) vs. 77.7% (95% CI: 70.5–83.6); specificity: 21.1% (95% CI: 19.5–22.9) vs. 44.4% (95% CI: 41.4–47.6); PPV 43.2% (95% CI: 37.5–48.9) vs. 47.5% (95% CI: 41.0–54.0); NPV 92.7% (95% CI: 85.2–99.9) vs. 84.2% (95% CI: 77.8–91.0). After repeating this analysis with 18 month cumulative ≥CIN3 as endpoint slightly higher sensitivity values were obtained (Table IV for the 33rd percentile threshold).
|Number ≥CIN2 positive (N)||Number ≤CIN1 positive (N)||Sensitivity (95% CI)||Specificity (95% CI)||PPV (95% CI)||NPV (95% CI)|
|HPV 16||93/103||105/153||88.2 (80.5–93.1)||33.1 (29.8–36.5)||55.9 (47.8–64.0)||74.7 (60.3–89.1)|
|HPV 18||10/11||37/47||86.5 (54.2–97.2)||20.6 (16.3–25.8)||25.0 (11.0–39.0)||83.3 (53.1–99.9)|
|HPV 31||17/18||68/91||95.3 (69.3–99.5)||26.3 (22.2–30.7)||24.1 (13.7–34.5)||95.8 (87.8–99.9)|
|HPV 33||16/18||16/26||90.6 (66.1–98.0)||44.2 (34.6–54.3)||60.5 (40.6–80.4)||83.3 (62.2–99.9)|
|All types combined||136/150||226/317||88.8 (82.7–92.9)||29.7 (27.5–32.0)||44.8 (38.9–50.7)||80.5 (70.7–90.3)|
|≥Borderline dyskaryosis||108/150||52/317||61.0 (53.6–67.9)||82.8 (79.3–85.8)||69.3 (64.3–73.9)||76.9 (73.2–80.2)|
|Number ≥CIN3 positive (N)||Number ≤CIN2 positive (N)||Sensitivity (95% CI)||Specificity (95% CI)||PPV (95% CI)||NPV (95% CI)|
|HPV 16||72/77||126/179||91.8 (83.3–96.2)||31 (28.0–34.3)||43.5 (35.4–51.6)||86.7 (75.6–97.8)|
|HPV 18||7/7||40/51||99.9 (0.0–100.0)||22.2 (17.6–27.7)||18.1 (5.6–30.6)||99.9 (0.0–100.0)|
|HPV 31||12/13||73/96||93.6 (61.0–99.3)||24.6 (20.8–28.6)||18.1 (7.9–26.3)||95.8 (87.8–99.9)|
|HPV 33||11/13||21/31||87 (56.5–97.2)||34.9 (26.9–43.8)||41.7 (21.4–62.0)||83.3 (62.2–99.9)|
|All types combined||102/110||260/357||91.7 (84.9–95.6)||28.1 (26.0–30.3)||33.7 (28.1–39.3)||89 (81.7–96.3)|
|≥Borderline dyskaryosis||87/110||73/357||71.2 (62.6–78.6)||80.4 (76.9–83.5)||59.1 (51.2–67.0)||87.6 (82.7–92.5)|
Furthermore, unlike our previous findings in women with abnormal cytology and underlying ≥CIN3,25 none of the thresholds combined for all 4 types revealed a 100% sensitivity for 18-month cumulative ≥CIN3. These findings may, in part, reflect the fact that women with normal cytology having or developing ≥CIN2 or ≥CIN3 demonstrate lower viral loads than those with abnormal cytology. Indeed, among those with ≥CIN2 who were HPV16-positive, the women with normal cytology at baseline had significantly lower viral loads than their counterparts with abnormal cytology (median 1.2 c/c (range <0.01–113) vs. median 9.1 c/c (range <0.01–1.2 × 104); median test p < 0.001). Similar results were obtained for HPV16-positive women with ≥CIN3 (median test; p < 0.001). The number of HPV18, -31 and -33 positive women with ≥CIN2/3 lesions was too low for such analysis. In summary, no viral load thresholds could be defined below which cumulative 18-month ≥CIN2 and ≥CIN3 could be excluded in HPV-positive women.
Clinical performance of viral load analysis relative to cytology
We further evaluated whether viral load assessment can be an alternative for or additive tool to cytology to stratify hrHPV-positive women for risk of high-grade CIN or worse. To that end, we performed ROC-curve analysis to compare the sensitivities and specificities of the viral load thresholds for the defined endpoints with that of cytology. For ≥borderline dyskaryosis (abnormal cytology) these values are also given in Table IV. At a similar sensitivity, the specificity for ≥CIN2 of cytology was always higher than that of the viral load thresholds (Fig. 2). Moreover, adding viral load testing to cytology in HPV DNA-positive women increased the sensitivity for ≥CIN2 at the cost of a marked decrease in specificity for ≥CIN2. Similar results were obtained when using ≥CIN3 as endpoint (data not shown). Altogether, these findings indicate that viral load assessment, despite its increased sensitivity has no additive value over that of cytology to stratify GP5+/6+-PCR-based hrHPV DNA-positive women for risk of ≥CIN2 or ≥CIN3 because it is too non-specific and consequently would lead to an increase in unnecessary referrals.34
Several population-based cervical screening trials9, 10, 35 have indicated that hrHPV DNA testing using a clinically validated assay12 is likely to replace cytology in primary cervical screening in the near future. Studies, however, also indicated that specificity of hrHPV testing remains subject to improvement.35, 36 In a population-based cervical screening cohort (i.e., POBASCAM), we therefore evaluated viral load assessment as a candidate biomarker to stratify GP5+/6+-PCR-based hrHPV-positive women for risk of high-grade CIN or worse within 18 months of follow-up (i.e., both the histologic endpoints ≥CIN2 and ≥CIN3 were evaluated). Increased viral loads, particularly of HPV 16, 31 and 33, were predictive of cumulative high-grade CIN or worse diagnosed within an 18-month period after enrollment. Our data are in line with those of previous studies suggesting that increased viral loads of most hrHPV types are associated with high-grade CIN lesions diagnosed up to 2 years of follow-up.21, 37, 38 On the other hand, the long-term predictive value for incident cancer precursor lesions diagnosed after many years following the baseline sample16, 39 seems to be limited to HPV-16 load solely.37 In agreement with previous observations,16, 17 we also found a tendency for a positive association of in-time increasing viral load with risk of high-grade CIN or worse.
However, despite the predictive value of viral load for 18 months ≥CIN2 or ≥CIN3 we could not translate our findings into a useful threshold value for risk stratification at the individual patient level since utilization of thresholds had a negative influence on the cumulative 18 months clinical sensitivity. Moreover, adjustment of the viral load thresholds to less than the 25th percentile revealed sensitivities for ≥CIN2 and ≥CIN3 that were lower than those previously reported.25 This was due to the wide range of overlapping viral load values in both groups of women with and without ≥CIN2 and ≥CIN3, particularly amongst women with normal cytology at baseline. Since additional viral load assessment could not enhance the clinical specificity of the GP5+/6+-PCR without a marked decrease in clinical sensitivity the analytical sensitivity of the GP5+/6+-PCR for the detection of hrHPV seems already to be at the most optimal level to ensure an ideal balance between clinical sensitivity and specificity for ≥CIN2.
Furthermore we found that viral load analysis had no alternative or additive value with respect to the clinical sensitivity and specificity for both ≥CIN2 and ≥CIN3 over that of cytology. It is noteworthy that in this study women with a double-negative cytology/hrHPV test result at 18 months were neither referred for colposcopy nor underwent any other follow-up examination and, consequently, ≥CIN2 lesions in this group might have been missed. However, from a previous study, we know that the 5 year ≥CIN2 risk of double negative women is only 0.1%,9 and since this group comprised only 56 women the number of potentially undiagnosed ≥CIN2 lesions would be negligible. Therefore, it can be concluded that in cervical screening settings where qualitative appropriate cytology is available as triage tool, viral load assessment has no extra value for stratification of hrHPV GP5+/6+-PCR-positive women for risk of ≥CIN2 or ≥CIN3. Furthermore, it is of note that this study does not have an arm in which women are referred on the basis of viral load levels. As a consequence, we cannot exclude the fact that a slightly higher PPV for viral load analysis would have been obtained when referring on the basis of viral load levels. Nevertheless, the current data do not give any evidence that viral load would perform better than cytology as a triage tool for HPV DNA positive women.
It is noteworthy that viral load analysis was only performed on samples with HPV loads above the assay threshold of the GP5+/6+-PCR assay. That other studies seemed to find stronger associations between viral load and ≥CIN2 might be explained by the fact that these studies most likely also included women with viral load levels below the detection limit of GP5+/5+-PCR, as a result of the use of a different, analytically more sensitive assay as test to distinguish hrHPV positive women.14, 21, 37 Since GP5+/6+-PCR-negative women have virtually no risk of developing ≥CIN29 it is obvious that inclusion of women with extremely low viral loads, undetectable by GP5+-6+-PCR, would result in higher associations.
Hc2 and GP5+/6+-PCR display similar clinical performances.40 It is, however, of note that the clinical specificity of hc2 is somewhat lower than that of GP5+/6+-PCR, but it increases and approaches that of GP5+/6+-PCR when the assay cut off is increased from 1 RLU/CO to 2–3 RLU/CO.6, 40, 41 This is partly due to a reduction in cross-reactivity of hc2 with low-risk HPV types at these increased cut-off values.42 Thus, additional viral load analysis is also unlikely to be of additive clinical value in case of a hrHPV hc2-positive test result, particularly when an assay cut-off of 2–3 RLU/CO were to be used.
Finally, we excluded women with multiple HPV infections that include HPV16, -18, -31 and/or -33 for viral load assessment, because it is impossible to resolve which HPV type in such infections would be responsible for the possible development of a lesion. The possible impact of including the multiple infections would have been that sensitivity and PPV for ≥CIN2 would to a certain extent be underestimated and specificity and NPV would be overestimated. However, in this case it would be unlikely that viral load would perform better than cytology as a triage tool for HPV DNA positive women.
Still, it is of note that the sensitivity and specificity values of cytology presented in this study can not be simply compared with those of other studies involving an unselected population,9, 35 because this study was limited to a group of GP5+/6+-PCR-based hrHPV-positive women, with single infections of either HPV16, -18, -31 or -33. Nevertheless, cytology harbors a rather low clinical sensitivity for ≥CIN2 in population-based screening. Therefore, management of all GP5+/6+-PCR-based hrHPV-positive, cytology-negative women is needed as these women still harbor a substantial risk for ≥CIN2.43 Thus, other means of specificity enhancement will be necessary for cervical screening to reduce the number of women with normal cytology who would undergo redundant clinical follow-up procedures based on a transient hrHPV infection. As of this moment, repeat testing of hrHPV-positive, cytology-negative women with both hrHPV testing, and cytology at 6–12 months seems to be the most effective strategy.43, 44 Other viral markers including genotyping,45, 46 analysis of type persistence and/or E6/E7 mRNA transcript analysis47 as well as host-cell biomarkers reflecting a transforming hrHPV infection-like p16INK4a overexpression48 warrant further investigation for their potential to enhance the efficacy of the follow-up.49 Additionally, research lines aiming at the identification of novel candidate host-cell markers for risk stratification of hrHPV-positive women remain of major importance.50, 51
In conclusion, despite the predictive value of viral load for cumulative high-grade CIN or worse, viral load assessment had no additive clinical value over cytology as a potential triage tool for hrHPV GP5+/6+-PCR-positive women. These findings argue for the need for other biomarkers eligible for risk stratification of hrHPV-positive women within a screening setting.
The authors thank Mr. Anthonie Groothuismink, Ms. Bistra Veleva, and Ms. Padamjit Kaur for excellent technical assistance.
- 37High 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., , , , , , , , , , , .