Discussion
This is the first study to assess the predictive potential of the S5
methylation classifier in a larger screening setting of patients with
abnormal results between HPV and cytology in an Asian country. S5
classifier, a multigene methylation test, which has shown good
performance in the United Kingdom30,32 , Canada33, and Mexico31,
Colombia34 and recently in a China colposcopy referral
population conducted by our team13. In this study, we
enrolled the screening population as study setting, and we evaluated the
predictive potential of S5 methylation for progression of specific
untreated squamous intraepithelial lesion.
Here, we validated the effectiveness of S5 classifier at the cutoff 2.85
in the present screening population setting. We validated a DNA
methylation classifier of HSIL+ histology, among women with abnormal
results between HPV and cytology from a China screening population. Our
study included 938 baseline samples (including HPV, cytology and S5
methylation) from women who were referred colposcopy at first screening
or had 2 years of active follow-up and culminating in colposcopy clinics
directed biopsy diagnoses if positive for either HPV or cytology tests.
To address potential bias in methylation levels by age at diagnosis or
time of follow-up, our controls were age and time to diagnosis matched
and randomly chosen among all baseline ≥HSIL+ women. We conducted an
independent analysis in the RF arm and IC arm in Table 2, although the
sensitivity and specificity of S5 is different, the result is still in
line with the overall conclusion (which means that the sensitivity and
specificity of S5 in both cohorts are better than the performance of
HPV16/18 and cytology). The genotyping and methylation assays, as well
as the verification of the histological diagnoses, were conducted
independently and blindly after the end of the study. To date, there are
few reports comparing the performance of cytology and HPV16/18
genotyping with DNA methylation assays to detect HSIL+ in women with
abnormal results between HPV and cytology. The study compared the
performance of S5 for HSIL+ detection to repeatedly cytology testing and
HPV genotyping in women with abnormal results between HPV and cytology
in developing countries setting, objectively.
The AUC obtained in this study was 0.80 (95% CI 0.74-0.85). At a
specificity set at 79.9% that corresponds to using S5 at a cut-off
value of 2.85, the sensitivity of S5 for HSIL+ were 76.1%, which were
significantly higher than the sensitivities of HPV16/18 (P =
0.039) or cytology (P < 0.001). In addition, S5
methylation classifier also outperformed the comparator triage testes
(HPV and cytology) and their combination regarding the specificity.
Considering that we are comparing different triage tests among women
with above the average risk for HSIL+ (cytology≥(TCT)HSIL or/and
hrHPV+), S5 acts as a good triage alternative since this test decreased
the false positive rate by near 35% (Table 4) and exhibited higher
sensitivity than HPV 16/18, cytology, or combination of these two tests
for both HSIL+ endpoints. This characteristic of S5 is especially
valuable for remote areas of developing countries, where higher
sensitivity is crucial to identify at-risk women in fewer screening
visits and decreasing the use of resources to follow-up women with low
risk of disease.
In our study, cytology had a specificity of 54.4%, but had the test
with by far the highest false negative rate. We recognized that the
sensitivity of our cytology was lower than the performance of cytology
in several studies in developed countries, but several studies have
demonstrated that the sensitivity in developing countries may be as low
as 30%. Although the selection bias about population occurs, it is
worth noting that despite many efforts to improve cytology quality and
sensitivity, difficulties in quality controlling and delays in diagnosis
still prevail. We used the predefined cutoff at UK population (S5 cutoff
= 0.8) that corresponded to previous studies with
S531,32, while the S5 at cutoff 0.8 (95.5%) vs the
sensitivity of S5 at cutoff 2.85 (76.1%), the specificity(19.4%, 95%
CI 17.8-21.7) at 0.8 cutoff is much lower than 79.9% at 2.85 cutoff.
Future work is planned to assess the performance of S5 as a triage to
determine appropriate cutoffs within a screening population. Also,
further work is needed to determine the performance of the S5
methylation assay in vaccinated populations.
Also, we analyzed the distribution of S5 panels scoring and host-cell
gene named EPB41L3 scoring among a screening population (Figure
2 and Figure 3 ), the results were in line with the distribution among
the colposcopy referral population. Additionally, the changes in the
level of methylation of S5 panels and host-cell gene named EPB41L3
compared if methylation of EPB41L3 was different among different
cytology results and different HPV status, the result showed very little
difference (P = 0.31) (Figure 4 ). This result indicates
that it is difficult to conduct a screening test only by measuring the
methylation levels EPB41L3, which was in line with some other previous
studies29,32.
What’s more, we reveal a new utility of the S5 DNA methylation
classifier measurement, specifically as a classifier not only for an
enlarged screening population, but also for assessing risk of
progression specific untreated squamous intraepithelial lesions. A
promising prognostic test for SIL could greatly improve the current
screening and even alter further treatment algorithms. A generally
acknowledged difficulty of current strategies lies in the subjectivity
and inter-observer variability in not only cytological but also
histological diagnoses. These limitations result in misclassification of
lesions, unnecessary colposcopy referrals, multiple follow-up visits,
and either delayed treatment or over-treatment, exacerbating the
potential harm to the patient meanwhile excessive burdening to medical
practice. An improved predictive test could optimize the current
management so that cases with progressive potential could be treated
sooner and cases with regressive potential perhaps be indicated
untreated for longer periods to allow more regressions. It should be
considered that the natural history of long-term HPV persistence with
respect to eventual true progression of abnormal screening results
(abnormal screening results between HPV and cytology) to ≥HSIL+ beyond 2
years remains unclear. According to the current Chinese clinical
algorithm, persistence of abnormal screening results (abnormal screening
results between HPV and cytology) for 2 years was an indication for
colposcopy referrals or even further treatment; however, we do not know
what proportion of these colposcopy referrals were necessary to prevent
cervical cancer.
In this study, of all the women (abnormal results between HPV and
cytology; meanwhile HPV16/18- and < (TCT)HSIL result of
cytology) progressed to ≥HSIL+ within 2 years, most of these cases were
positive for the S5 methylation classifier (S5 methylation classifier ≥
2.85) at the first visiting. All the cases with HPV16/18 positive or
cytology ≥ (TCT)HSIL+ were referred to immediate colposcopy according to
the current Chinese algorithms, thus it is not clear the proportion of
progressed women with HPV16/18 positive or cytology ≥ (TCT)HSIL+, and it
is plausible that further prospective study is needed to demonstrate the
issue. Even so, these results fully indicate that S5 methylation
classifier with great potential to predict the progression of women with
abnormal results between HPV and cytology.
According to the call for “action to eliminate cervical cancer” issued
by the World Health Organization in 20199, increasing
effective screening of women with hrHPV testing is necessary. However,
it is difficult to implement the HPV DNA testing frequently for
populations living in remote areas where there few opportunities to
screen women at proper intervals and for follow-up after
screening4,35. What’s more, most HPV infections are
transient so immediate ablative treatments can lead to unnecessary
gynecological harms for women with low risk of
disease36. For this reason, primary screening with HPV
testing and/or cytology requires other triage tests to identify women at
high-risk of high-grade disease among those who are with abnormal
results between HPV and cytology. Considering recalling women for a
second test after screening is challenging or impossible in developing
countries; therefore, risk triage should ideally occur at the screening
visit in these settings. Hence, these settings have unique needs for
triage strategies.
Considering these results, it is possible to envisage a screening test
that simultaneously the current tests (HPV genotyping and cytology) and
measurement of S5 methylation. As depicted in Figure 8 , such
potential screening strategy would provide the benefit of immediate and
more accurate results for triaging women as below: (I) negative for all
biomarkers, who would go back to routine screening; (II) referred to
colposcopy (according to the current algorithm: HPV16/18 positive or
cytology ≥ (TCT)HSIL+ or repeated abnormal result among HPV and
cytology) and methylation positive, would be given priority for
colposcopy inspection than who referred colposcopy and methylation
negative, and (III) repeated follow up (according to the current
algorithm) and methylation positive, would be given priority for
follow-up at proper intervals than who were repeated follow up and
methylation negative, in developing countries especially some remote
areas.
The strength of this study is the validation of the S5 classifier in a
routine screening study in the China with blinding of all results to the
lab technicians, and the use of prespecified cutoffs for the methylation
classifiers which minimized the risks for bias and overfitting. In
practice, hrHPV-positive women could have the methylation tests
performed on the original samples in a reflex manner, triaging women at
risk to colposcopy and thereby reducing anxiety and overtreatment in the
low-risk women. Possible concerns over missing some of the LSIL and HSIL
might be addressed by referring women negative or low risk by the DNA
methylation classifier to repeat HPV testing and cytology in proper
follow-up duration.
A weakness is that our study was restricted to the length of follow-up
short, which may result in some bias. While potential statistical bias,
it is under the circumstance of the current algorithm in China based on
the guideline: all the women (with HPV16/18 positive and/or cytology
results ≥(TCT)HSIL+) were advised to refer colposcopy. Thus, it is
difficult to arranged women for immediate colposcopy or follow up
randomly, also difficult for RF cohort undergo colposcopy before the
following-up. Considering that in this study we focused the screening
population (with abnormal screening results) at the first visiting, IR
cohort and RF cohort were both included. Since the RF cohort including
women with (“HPV positive except HPV16/18” and “abnormal cytology
results except ≥(TCT)HSIL”), who were followed-up every 6 months for 2
years follow-up period, and need 2 years following up period to
determine whether to be enrolled in, the current study (including IC
cohort and RF cohort) lasted for a long time to complete the data
collection. However, a larger population is needed in the further study.
In addition, our results might not be directly generalized to other
histological diagnosis (diagnosis using cervical intraepithelial
neoplasia in some studies) properly. Despite the differences in study
designs and in the proportion of the HSIL+ cases be included, our
results are remarkably similar to other
studies30,32,33. The S5 methylation test accurately
identified women with higher risk of cervical high-grade disease and
cancer among those who with abnormal result between HPV and cytology.
Even further, our study demonstrated that S5 outperforms both cytology
and HPV16/18 in detecting HSIL+ in Chinese women. Currently, S5 DNA
methylation test is labor intensive and costly. Based on the practice of
our study13, the cost of an S5 test is not higher than
that of a routine HPV screening and reflex genotyping, and it could be
balanced against the expected 30 to 50% reduction in colposcopy
referral costs. The recent developments of affordable and scalable
next-generation sequencing assays strengthen our proposal that the S5
DNA methylation classifier test may be an acceptable strategy for the
triage of women with abnormal result between HPV and cytology in
developing countries. Strategies combing screening and immediate
clinical management are urgently needed for developing countries
settings where infrastructure for follow-up after screening is limited.
S5 might be a potential classifier in developing countries for reducing
costs and encouraging doctors to focus on the women at real risk of
cervical cancer. Our results warrant further clinical validation of S5
in larger prospective population-based screening trials.