The impact of human papillomavirus genotype on colposcopic appearance: a cross-sectional analysis

Authors


Abstract

Objective

To study colposcopic performance in diagnosing high-grade cervical intraepithelial neoplasia or cervical cancer (CIN2+ and CIN3+) using colposcopic characteristics and high-risk human papillomavirus (hrHPV) genotyping.

Design

Cross-sectional multicentre study.

Setting

Two colposcopy clinics in The Netherlands and Spain.

Population

Six hundred and ten women aged 17 years and older referred for colposcopy because of abnormal cytology.

Methods

A cervical smear was obtained. Colposcopists identified the worst lesion, graded their impression and scored the colposcopic characteristics of the lesions. Up to four biopsies were collected, including one biopsy from visually normal tissue.

Main outcome measures

CIN2+ and CIN3+, positive for HPV16 or other high-risk HPV types (non-16 hrHPV-positive).

Results

The mean age in HPV16-positive CIN2+ women was 35.1 years compared with 39.1 years in women with other hrHPV types (= 0.002). Sensitivity for colposcopy to detect CIN2+ was 87.9% (95%CI 83.2–91.5), using colposcopic cut-off of ‘any abnormality’. The remaining CIN2+ were found by a biopsy from visually normal tissue or endocervical curettage (ECC). Detection of CIN2+ by lesion-targeted biopsies was not different between HPV16-positive women [119/135; 88.1% (95%CI 81.2–92.9)] and non-16 hrHPV-positive women [100/115; 87.0% (95%CI 79.1–92.3); = 0.776]. In multivariate analysis, ‘acetowhitening’ [odds ratio (OR) 1.91, 95%CI 1.56–3.17], ‘time of appearance’ (OR 1.95, 95%CI 1.21–3.15) and ‘lesion >25% of visible cervix’ (OR 2.25, 95%CI 1.44–3.51) were associated with CIN2+.

Conclusions

In this population following European screening practice, HPV16-related CIN2+ lesions were detected at younger age and showed similar colposcopic impression as non-16 hrHPV high-grade lesions. There was no relationship between any of the colposcopic characteristics and HPV16 status.

Introduction

Almost all invasive cervical cancer (ICC) cases are caused by persistent infections with high-risk human papillomavirus (hrHPV) through premalignant stages of cervical intraepithelial neoplasia (CIN).[1, 2] Cervical cytology screening programmes have improved the detection of precancerous lesions and, as a result, cervical cancer incidence has decreased in developed countries.[3] Women with abnormal cytology are offered colposcopy, which is the standard for identifying CIN2+ (CIN2, CIN3 and ICC) in most countries. Current biopsy procedures rely on visual features to identify the area on the cervix that most likely represents the worst lesion.[4-8] However, colposcopy has a sensitivity of 50–70% to detect CIN2+.[9-13] Gains in sensitivity can be obtained by increasing the number of biopsies.[14] Moreover, in a study performing random biopsies of regions without abnormalities, 23–37% of the overall CIN2+ lesions were detected in this biopsy only.[15]

The risk of developing ICC within the group of hrHPV genotypes varies substantially. HPV16 is found to be the genotype with the highest carcinogenic potential; it is associated with the highest risk of the development of CIN3 and ICC.[16-19] Also, HPV16-related CIN2+ has been found at young age.[20, 21] HPV genotype 16 is suggested to cause more definite visual abnormalities than other HPV types.[5, 22] Women with CIN2+ frequently present with multiple HPV genotypes in cytology.[23] In a study using laser capture microdissection (LCM) in women positive for multiple HPV types in cytology we have shown that HPV16 is the most predominant causal genotype in CIN2+ lesions.[24] This suggests that etiologically, HPV16 has an even more important role than previously thought. Although a growing interest exists for clinical use of hrHPV testing to triage women who have minor cytological changes and in primary screening, HPV genotyping is currently not implemented as standard clinical practice in most countries.[25]

The objective of this study was to evaluate the visual appearance of the cervix using colposcopic characteristics combined with HPV genotyping to predict CIN2+ in women who were referred for colposcopy.

Methods

Study design and population

The EVAH-study (Evaluating the Visual Appearance of cervical lesions in relation its histological diagnosis, Human papillomavirus genotype and other viral parameters) is a multicentre study conducted between August 2010 and October 2012. In total, 610 women aged 17 years and older who had an abnormal Pap smear result defined as atypical cells of undetermined significance (ASC-US) or worse (disregarding HPV status), who were referred for colposcopic evaluation to Reinier de Graaf Groep, Voorburg, The Netherlands or to Hospital Clínic, Barcelona, Spain, were included.

Patients were excluded if they had had previous treatment of cervical pathology, had a confirmed diagnosis of ICC at the time of referral, had insufficient knowledge of the Dutch or Spanish language or were pregnant or breast-feeding at the date of colposcopy or 3 months before. This study has received approval from the medical ethical board at both hospitals. Informed consent was obtained from all participating women.

Cytology, colposcopy and specimen collection

Before colposcopy, a liquid-based cytology sample using a Cervex-Brush® (Rovers Medical Devices BV, Oss, The Netherlands) was obtained. The Cervex-Brush was rinsed in ThinPrep® medium (Hologic, Marlborough, MA, USA) at Hospital Clínic, and in SurePath™ medium (Klinipath BV, Duiven, The Netherlands) at Reinier de Graaf Groep. Cytological examination and classification were performed at the local laboratory according to the Bethesda 2001 classification in Spain (negative, ASC-US, low-grade squamous intraepithelial lesion (LSIL), high-grade squamous intraepithelial lesion (HSIL), atypical glandular cells of undetermined significance (AGUS), atypical squamous cells, cannot exclude HSIL (ASC-H), adenocarcinoma in situ (AIS) or ICC). In The Netherlands grading was done according to the CISOE-A (composition, inflammation, squamous epithelium, other and endocervical columnar epithelium, and adequacy of the smear), which was translated to the Bethesda 2001 classification as described earlier.[26]

Nine colposcopists were involved in this study, seven expert and two junior colposcopists with supervision from an expert colposcopist. Five percent acetic acid was used to elicit the acetowhite epithelial response. Grading of visual impression was done as normal (including squamous metaplasia), CIN1, CIN2, CIN3 (including AIS) or ICC and the impression was coded in the final data analysis as normal (N), low-grade (LG; CIN1) or high-grade (HG; CIN2+, including ICC). Colposcopists were asked to score the lesion margins (geographical, smooth, internal borders), colour of the lesions (absent, translucent, intermediate white, opaque white), punctation (absent, fine, coarse), mosaicism (absent, fine, coarse), the presence of atypical vessels, time of appearance (slow, fast) and size of the lesions (0, <25, 25–50, >50% of the cervix). This scoring system was based on the Reid colposcopy index designed by Reid and Scazi.[27] In this study, multiple biopsies during colposcopy were collected. Up to four biopsies were taken from different lesions, or distinct areas within a large complex lesion were biopsied separately. If fewer than four directed biopsies were taken, a biopsy from visually normal tissue was added. Using this refined biopsy protocol, we aimed to detect more underlying CIN2+ than with the current approach, in which usually only the most severe lesion is identified and biopsied. Endocervical curettage (ECC) was performed in cases where the squamocolumnar junction (SCJ) was not completely visualized. In accordance with the national guidelines in The Netherlands and Spain, histologically confirmed high-grade lesions diagnosed as CIN2+ were treated.

Histological processing

Individual biopsy specimens, including ECC samples when taken, were processed separately, fixed in 10% neutral buffered formalin and paraffin-embedded. Haematoxylin and eosin (H&E) sections were examined by a local pathologist and classified as negative (including squamous metaplasia), CIN1, CIN2, CIN3 (including AIS) or ICC. All biopsies were independently reviewed by a second gynaecological pathologist. In case of disagreement between the original and review diagnosis, a third pathologist reviewed the discordant cases independently. Consensus diagnosis was determined by the agreement of two of three interpretations. All pathologists were blinded for HPV status. The final histological diagnosis per case was based on the worst diagnosis found in all specimens of each woman.

hrHPV genotyping

DNA extraction was performed using 250 μl of the cytology specimen to obtain 100 μl of eluate with the QIAamp MinElute Virus Spin kit (QIAgen Inc., Valencia, CA, USA). GP5+/6+ PCR-based HPV genotyping was performed at both laboratories in The Netherlands and Spain. In The Netherlands, 10 μl of isolated DNA was amplified and genotyped using the LMNX HPV GP Genotyping kit (Labo Bio-medical Products BV, Rijswijk, The Netherlands) as described by the manufacturer.[28] The LMNX test utilizes Luminex xMAP high-throughput technology to identify 18 HPV types; i.e. HPV16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73 and 82. In Spain, 10 μl of isolated DNA was amplified by the GP5+/6+ PCR and hrHPV was detected by the EIA (Diassay, Rijswijk, The Netherlands) according to the manufacturer's instructions. With the GP5+/6+ PCR–EIA test, 14 hrHPV types can be targeted: HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68. Next, the EIA-positive GP5+/6+ amplimers were genotyped using the Genotyping kit HPV GP (Diassay), according to the manufacturer's instructions. The strip-based genotyping test targets the same 18 HPV types as the LMNX test. To compare the performance of the two genotyping tests, the first 60 cytology samples included in the study were analysed with both HPV genotyping methods. The agreement between both methods in hrHPV detection was 86.7% (= 0.727, kappa= 0.72, 95% CI 0.54–0.90). Of the 32 samples that were hrHPV-positive in both methods, 20 (63%) showed identical hrHPV genotypes, 10 (31%) concordant hrHPV genotypes and 2 (6%) discordant genotypes.

Statistical analysis

Independent sample t-tests and chi-square tests were used to compare age, age at sexual debut, cytology diagnosis and hrHPV status. Analyses were stratified by hrHPV status and HPV16 status. HPV detection and genotype analyses were restricted to 14 hrHPV types (HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 and 68). HPV16 status was defined as positive if HPV16 was detected by GP5+/6+ in the cytology specimen. Positivity for any other hrHPV type present except HPV16 was defined as non-16 hrHPV-positive.

Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for colposcopic performance were calculated. As a cut-off for abnormal colposcopy, both low-grade or worse visual impression (any grade of abnormality) and high-grade or worse visual impression (HG+) were used. The association between lesion size, HPV16 status, and age in women with CIN2+ or CIN3+ was examined using linear logistic regression. Level of statistical significance was set at 0.05. Colposcopic characteristics were studied on the cervix level: when multiple lesions with different colposcopic characteristics were present, the worst colposcopic features were used for analysis. Contingency tables using the Mann–Whitney U-test and Pearson's chi-square statistics were used to analyse the association between the different colposcopic characteristics and CIN2+/CIN3+ diagnosis and the association between HPV16 status and the different features. Odds ratios (ORs) and 95% CI were calculated. In binary logistic regression, the association between CIN2+ and colposcopic characteristics was studied, adjusted for age. Colposcopic characteristics were included in the model by forward selection using the Wald test-statistic as selection criterion. The selection was stopped when the P-value of the regression coefficient was above 0.05. Analyses were performed using SPSS version 20.0 for Windows (Chicago, IL, USA).

Results

Baseline characteristics

All women with an abnormal cytology result and no exclusion criteria who were referred between August 2010 and October 2012 were included in the study (n = 610). Clinical characteristics of the study group are described in Table S1. The mean age of the women was 36.5 years (SD 10.9) but the age was slightly different between the study centres (Barcelona: 36.0 years and Voorburg: 39.3 years; = 0.002). The mean age of sexual debut was 18 years (SD 2.8). Referral cytology diagnoses were: ASC-US (= 71; 11.6%), LSIL (= 200; 32.8%), HSIL (= 319; 52.3%), ASC-H (= 18; 3.0%), AGUS (= 1; 0.2%) and adenocarcinoma (= 1; 0.2%). Cytology results at enrolment are shown in Table S1. HrHPV positivity at enrolment was not significantly different between the two sites (= 0.510); 461 women (75.6%) were positive for hrHPV. Overall, 221 (36.2%) women had a histologically negative diagnosis, 123 (20.2%) women were diagnosed with CIN1, 144 (23.6%) with CIN2 and 122 (20.0%) with CIN3+ (including six with ICC). For two women the HPV genotyping data were not available and these women were excluded from further analysis.

HPV genotyping results

The HPV genotype distribution among the 461 hrHPV-positive women was as follows; HPV16 (= 193; 41.9%), HPV31 (= 73; 15.8%), HPV66 (= 43; 9.3%), HPV18 (= 42; 9.1%), HPV56 (= 39; 8.5%), HPV51 (= 37; 8.0%), HPV33 (= 29; 6.3%), HPV52 (= 29; 6.3%), HPV58 (= 26; 5.6%), HPV39 (= 22; 4.8%), HPV45 (= 19; 4.1%), HPV59 (= 16; 3.5%), HPV35 (= 15; 3.3%), HPV68 (= 11; 2.4%). Additionally, the following HPV types were detected; HPV73 (= 10; 2.2%), HPV53 (= 6; 1.3%), HPV82 (= 5; 1.1%), HPV26 (= 1; 0.2%). HrHPV was detected in 211/343 (61.5%) women with a histological diagnosis of negative or CIN1 and in 250/265 (94.3%) of women with CIN2+ (< 0.001). Both single and multiple infections were detected and HPV genotype distribution was calculated per individual HPV type. Multiple hrHPV infections were detected in 116/461 (25.2%) of the hrHPV-positive women. Fifty-seven of 211 (27.0%) hrHPV-positive women with a diagnosis of negative or CIN1 versus 59/250 (23.6%) of women with CIN2+ had a multiple hrHPV-type infection (= 0.45).

Age and HPV16 status

The mean age in the HPV16-positive women with CIN2+ and CIN3+ was significantly lower compared with the non-16 hrHPV-positive cases; the mean age for HPV16-positive women with CIN2+ was 35.1 years and for non-16 hrHPV-positive women 39.1 years (= 0.002). The mean age for HPV16-positive CIN3+ cases was 36.4 years, versus 40.5 years in the non-16 hrHPV-positive group (= 0.048). Also, there was a significant shorter sexual activity span between HVP16-positive women (17.2 years) with CIN2+ lesions compared with non-16 hrHPV-positive women (21.3 years; = 0.001). An equal trend was found for CIN3+ cases, although not significant (18.6 years versus 22.2 years; = 0.068). No difference in number of lifetime partners was found between the two groups (= 2.7; = 0.701). In linear regression, age in women with CIN2+ was associated with HPV16 status (HPV16-positive women; estimate 3.7 years younger, 95% CI −6.2 to −1.3, = 0.003). The association remained significant after adjusting for lesion size categorized as <25, 25–50% or >50% of the cervix (HPV16-positive women; estimate 3.7 years younger, 95% CI −6.1 to −1.3, = 0.003). For women with CIN3+ only, the univariate association between age and HPV16 was also significant (HPV16-positive women; estimate 4.1 years younger, 95% CI −8.0 to −0.2, = 0.040), and the association adjusted for lesion size was nearly significant (HPV16-positive women; estimate 3.7 years younger, 95% CI −7.7 to 0.2, = 0.065).

Colposcopic performance and HPV16 status

To evaluate the colposcopic performance we studied the relationship between colposcopic impression and histological outcome per HPV group (total hrHPV-positive, HVP16-positive and non-16 hrHPV-positive) (Tables 1 and S2). In total, 233/265 (87.9%) CIN2+ cases and 108/122 (88.5%) CIN3+ cases had an abnormal colposcopic impression (any grade of abnormality). There was an impression of HG or ICC (HG+) in 163/265 (61.5%) CIN2+ cases and in 87/122 (71.3%) CIN3+ cases (Table 1). Five of six ICC were recognized as ICC by the colposcopists. In one woman with adenocarcinoma as the cytological referral diagnosis, the SCJ was not visualized and an ECC was performed which showed ICC. In HPV16-positive women, 119/135 (88.1%) CIN2+ cases had a colposcopic impression of any grade of abnormality compared with 100/115 (87.0%) CIN2+ cases in non-16 hrHPV-positive women (= 0.776). Sensitivity for detection of CIN3+ in HPV16+ versus non-16 hrHPV-positive cases was 91.8 versus 82.2%, respectively (= 0.119) (Table 1). When HG+ visual impression was used as cut-off for abnormal colposcopy, sensitivity was not significantly different either (= 0.279 for CIN2+ and = 0.332 for CIN3+). Previous studies have shown that multiple HPV infections often show complex and confluent multiple lesions.[24, 29, 30] Therefore calculations were limited to cases with single HPV genotype infections only (single HPV16+; = 131 versus single non-16 hrHPV-positive; = 215) and revealed no differences in colposcopic detectability for CIN2+ and CIN3+. The sensitivity for CIN2+ detection with any grade of abnormal impression in single HPV16-positive and single non-16 hrHPV-positive women was not significantly different: 88/100 (88.0%) versus 77/92 (83.7%); = 0.393. Also, there was no difference in sensitivity for CIN3+ detection [52/58 (89.7%) versus 28/36 (77.8%); = 0.118, respectively].

Table 1. Sensitivity and specificity of colposcopy for the detection of CIN2+ and CIN3+ lesions stratified by HPV status
Visual impressionCIN2+CIN3+
SensitivitySpecificitySensitivitySpecificity
n %95% CI n %95% CI n %95% CI n %95% CI
  1. Results are stratified for the visual impression of low-grade cervical intraepithelial neoplasia (CIN) or worse (any abnormality) versus high-grade CIN or worse (high-grade).

  2. a

    Including ICC.

Overall
Any abnormality233/26587.983.2–91.5152/34344.339.0–49.8108/12288.581.2–93.4170/48635.030.8–39.4
High-gradea163/26561.555.3–67.3282/34382.277.7–86.087/12271.362.3–79.0349/48671.867.5–75.7
hrHPV-positive
Any abnormality219/25087.682.7–91.390/21142.735.9–49.6104/11888.180.6–93.1107/34331.226.4–36.4
High-gradea154/25061.655.2–67.6175/21182.977.0–87.683/11870.361.1–78.2236/34368.863.6–73.6
HPV16-positive
Any abnormality119/13588.181.2–92.920/5834.522.8–48.267/7391.882.4–96.630/12025.017.7–33.9
High-gradea79/13558.549.7–66.848/5882.870.1–91.049/7367.155.0–77.480/12066.757.4–74.8
non-16 hrHPVpositive
Any abnormality100/11587.079.1–92.370/15345.837.7–54.037/4582.267.4–91.577/22334.528.4–41.2
High-gradea75/11565.255.7–73.7127/15383.075.9–88.434/4575.660.1–86.6156/22370.063.4–75.8
HPV negative
Any abnormality14/1593.366.0–99.762/13247.038.3–55.84/410039.6–10063/14344.135.8–52.6
High-gradea9/1560.032.9–82.5107/13281.173.1–87.14/410039.6–100113/14379.071.3–85.2

To investigate further the role of HPV16 in CIN2+ and CIN3+, the PPV of colposcopy in HPV16-positive versus non-16 hrHPV-positive women in this population was calculated (Table S3). PPV was higher in HPV16-positive women with CIN2+ than in non-16 hrHPV-positive women, for both any grade of abnormality and HG+ impression (75.8 versus 54.6%; < 0.001, and 88.8 versus 74.3%; = 0.011, respectively). This difference was also significant in CIN3+ cases (any grade of abnormality: 42.7 versus 20.2%; P < 0.001, and HG+: 55.1 versus 33.7%; = 0.003). We observed similar results for each of the nine colposcopists involved in the study.

Colposcopic characteristics and HPV16 status

In 429/608 (70.6%) cases, the SCJ was visible during colposcopy, at least one lesion-targeted biopsy (119 women had a random biopsy only) was collected and no missing colposcopic characteristics were reported. These cases were included in the analysis of colposcopic characteristics. ‘Time of appearance’ data were available for 422 women and ‘size of lesion’ for 423 women (Table 2). CIN2+ lesions had more distinct margins (= 0.001), more distinct acetowhitening (< 0.001), coarser punctation (< 0.001), coarser mosaic (= 0.02), appeared faster (< 0.001) and had a larger lesion size (< 0.001) than negative or CIN1 lesions. In addition, women with CIN3+ lesions had atypical vessels more often (= 0.001) than did women with negative tissue, CIN1 or CIN2. CIN3+ lesions did not show coarser mosaic (= 0.092). Atypical vessels were present in four of five women diagnosed with ICC and a visible SCJ. The time of appearance was fast and the lesion size was >50% in all ICC with SCJ visible. Mosaic, punctation and colour had a variable score in these cases (Table 2A).

Table 2. Different colposcopic characteristics in CIN2+ and CIN3+ cases (A) and in CIN2+ and CIN3+ cases stratified by HPV16 status (B). In total, 429 (70.6%) women with a visible transformation zone, at least one lesion-targeted biopsy taken and no missing reported characteristics were studied
CharacteristicsHistology
<CIN2CIN2+P-value<CIN3CIN3+P-value
(A)
Margins
Geographical95820.001144330.003
Smooth91134 16263
Internal borders720 1512
Colour
Absent96<0.001132<0.001
Translucent8444 11414
Intermediate85128 15657
Opaque1558 3934
Punctation
Absent144135<0.00123148<0.001
Fine4569 7737
Coarse432 1422
Mosaic
Absent1321390.020208630.092
Fine4661 8522
Coarse1536 2922
Atypical vessels
Absent1872220.168313960.001
Present614 911
Time of appearance a
Slow12382<0.00117728<0.001
Fast68149 14077
Size lesion b
0%54<0.00190<0.001
<25%10665 15120
25–50%4373 8630
>50%3592 7255
CharacteristicshrHPV-positives
CIN2+CIN3+
HPV16-positivenon-16 hrHPV-positiveP-valueHPV16-positivenon-16 hrHPV-positiveP-value
  1. a

    = 422. Missing data = 7.

  2. b

    = 423. Missing data = 6.

 (B)
Margins
Geographical44320.59220110.829
Smooth6759 4021
Internal borders119 75
Colour
Absent330.817200.813
Translucent2516 104
Intermediate6357 3322
Opaque3124 2211
Punctation
Absent72530.45431150.656
Fine3333 2214
Coarse1714 148
Mosaic
Absent65640.20036240.191
Fine3820 148
Coarse1916 175
Atypical vessels
Absent114940.86559340.545
Present86 83
Time of appearance a
Slow48300.1671990.777
Fast7267 4826
Size lesion b
0%310.180000.932
<25%3721 164
25–50%3535 1315
>50%4642 3717

The associations between colposcopic characteristics and CIN2+ (and CIN3+) were expressed by Mantel–Haenszel ORs, corrected for age. For end-point CIN2+, statistically significant associations were found for most characteristics but not for ‘atypical vessels’ or ‘mosaicism (absent versus present)’ (Table 3A). For end-point CIN3+, the presence of atypical vessels was significant and the presence of ‘mosaicism (absent versus present)’ was not significant (Table 3B). Colposcopic characteristics were also compared between HPV16-positive and non-16 hrHPV-positive CIN2+ and CIN3+ cases. There was no significant relationship between any of the colposcopic characteristics and HPV16 status, either for CIN2+ or for CIN3+ (Tables 2B and 3A,B). In a binary logistic regression, the association between CIN2+ and multiple colposcopy characteristics were studied, adjusted for age. Significant independent associations with CIN2+ were found for ‘acetowhitening’, ‘time of appearance’ and ‘lesion size (>25% of visible cervix)’. The corresponding ORs were 1.91 (95% CI 1.56–3.17), 1.95 (95% CI 1.21–3.15), and 2.25 (95% CI 1.44–3.51), respectively.

Table 3. Odds ratios (OR) for the different colposcopic characteristics in all women with CIN2+ and in women with CIN2+ stratified by HPV16 status (A) and ORs for women with CIN3+ and women with CIN3+ stratified by HPV16 status (B). All ORs are corrected for age
Colposcopic characteristicsTotal CIN2+HPV16-positive versus non-16 hrHPV-positive in CIN2+
OR (95% CI)OR (95% CI)
(A)
Margins (geographical vs smooth/internal borders)1.75 (1.18–2.61)0.77 (0.43–1.38)
Acetowhitening (absent/translucent versus intermediate/opaque)3.25 (2.12–4.99)0.70 (0.34–1.44)
Punctation (absent versus present)1.98 (1.29–3.03)0.78 (0.45–1.37)
Punctation (absent/fine versus coarse)6.74 (2.30–19.78)1.07 (0.49–2.33)
Mosaic (absent versus present)1.48 (0.98–2.26)1.26 (0.70–2.25)
Mosaic (absent/fine versus coarse)1.98 (1.03–3.81)0.93 (0.43–1.97)
Atypical vessels (absent versus present)1.50 (0.56–4.07)1.64 (0.51–5.26)
Time of appearance (slow versus fast)3.26 (2.17–4.91)0.59 (0.32–1.09)
Lesion size (cut-off 25%)3.24 (2.16–4.88)0.58 (0.30–1.10)
Lesion size (cut-off 50%)2.53 (1.62–3.96)0.90 (0.51–1.57)
Colposcopic characteristicsTotal CIN3+HPV16-positive versus non-16 hrHPV-positive in CIN3+
OR (95% CI)OR (95% CI)
(B)
Margins (geographical versus smooth/internal borders)1.79 (1.12–2.85)1.01 (0.42–2.42)
Acetowhitening (absent/translucent versus intermediate/opaque)3.51 (1.97–6.28)0.79 (0.22–2.90)
Punctation (absent versus present)2.93 (1.86–4.63)0.97 (0.41–2.27)
Punctation (absent/fine versus coarse)5.16 (2.52–10.55)1.40 (0.49–3.99)
Mosaic (absent versus present)1.32 (0.83–2.09)1.78 (0.72–4.37)
Mosaic (absent/fine versus coarse)2.69 (1.46–4.95)2.31 (0.72–7.37)
Atypical vessels (absent versus present)3.71 (4.47–9.40)1.84 (0.46–7.30)
Time of appearance (slow versus fast)3.43 (2.11–5.59)1.10 (0.40–3.06)
Lesion size (cut-off 25%)4.02 (2.37–6.80)0.64 (0.21–1.96)
Lesion size (cut-off 50%)3.54 (2.22–5.63)1.75 (0.75–4.12)

Discussion

Main findings

This study in women referred for colposcopy due to abnormal cervical cytology reveals that the colposcopic visual appearance of HPV16-related CIN2+ is not different from other hrHPV-type related lesions. The women were referred according to conventional European screening practice with a mean age of over 35 years and on the basis of any abnormal cytology grade including a single ASC-US smear. HPV testing was not part of the screening or triage procedure. No difference was observed in either sensitivity or specificity of colposcopic impression or individual colposcopic criteria for CIN2+, when caused by HPV16 or other hrHPV types. However, we confirmed that HPV16-positive women had a younger age at diagnosis of CIN3.[21, 31] We observed this effect for both CIN2+ and CIN3+ separately. Moreover, we found a significantly shorter span of sexual activity among women with HPV16-related CIN2+ than other carcinogenic HPV types, although these women had a similar number of lifetime partners. A similar, but non-statistically significant, trend for CIN3+ women was found. These findings suggest that HPV16-related CIN2+ lesions appear to develop faster than those related to other hrHPV types. When stratifying for age, we found no significant association between lesion size and HPV16 positivity. This implies that CIN2+ is detected at the same size irrespective of genotype but HPV16-related CIN2+ lesions reach the detection threshold faster than lesions related to other types because they are found at younger age.

In the present study, the sensitivity for detecting CIN2+ was 87.9% for any grade of abnormal impression and 61.5% for HG+ impression. In CIN3+ the sensitivities were 88.5 and 71.3% for any grade of abnormal impression and HG+ impression, respectively. When studying performance of different colposcopic characteristics irrespective of HPV status, all characteristics except for ‘presence of atypical vessels’ and ‘mosaicism (absent versus present)’ were significantly associated with CIN2+. The highest correlations were found for ‘acetowhitening’, ‘time of appearance’ and ‘lesion size of >25% of visible cervix’. In general, colposcopy has a sensitivity of 50–70% to detect high-grade lesions and ICC,[9-12] but the sensitivity can be increased by raising the number of collected biopsies.[14] The use of an extended biopsy protocol in our study with multiple targeted biopsies from all abnormal areas and a biopsy from visually negative tissue could explain the high sensitivity we have found. Unlike measures like sensitivity and specificity, the PPV depends on the prevalence of CIN2+ in the study population. We found a PPV of 72.8% for CIN2+ using colposcopic impression of HG+, representative for this population and broadly comparable to a recent UK study.[32]

Strengths and limitations

The main strength of this study is the large well described European study population representative of women with abnormal smears referred for colposcopy. In The Netherlands there is an organized 5-yearly screening program starting at the age of 30. In Spain there is opportunistic screening. In both countries the median age of women referred for colposcopy was over 35 years. A large number of women had CIN2+ as a histological endpoint. Various colposcopists from two different centres took part. We used a refined protocol of collecting multiple biopsies. Histological diagnosis of all biopsies was done by consensus diagnosis including expert pathologists and local pathologists who were blinded to HPV status at the time of diagnosis.

A limitation of this paper is that it is a cross-sectional study and loop electrosurgical excision procedure (LEEP) outcomes were not included as a histological endpoint. Hence, we lack longitudinal HPV genotyping and histological end point data. Different test algorithms were used for HPV genotyping in the two study centres. A pilot study performed on the first samples included in the study showed good agreement between HPV positivity rates and HPV genotypes. Furthermore, previous studies demonstrated a high genotyping agreement between both test algorithms.[28, 33] As in most clinical and epidemiological studies, HPV genotyping was performed on cytology specimens. In multiple infections, it is uncertain which genotype causes the lesion. Without techniques to define the presence of different HPV types in pre-neoplastic cells they may be incorrectly associated with CIN. Therefore, LCM combined with sensitive PCR (LCM-PCR) is currently being applied to study HPV type attribution on lesion level in this study.

Interpretation

Our findings are in contrast with Jeronimo et al.[5] who suggested that HPV16 causes more definite visual abnormalities than other HPV types, regardless of eventual histological diagnosis. The mean age of the study population of Jeronimo et al. was substantially lower than that of our study population (24 years versus 36.5 years). An explanation for the difference in findings between the studies is that high-grade cervical precancer has to grow to a certain size before it becomes detectable during colposcopy and that high-grade lesions evolve more rapidly in women with HPV16 infections than many other hrHPV types. This is consistent with evidence that ICC associated with HPV16 occurs at an earlier age than that associated with many other hrHPV types and that the development of ICC from CIN3 is associated with a large lesion size.[18, 34] Lesions driven by other hrHPV types than HPV16 in the older women in the present study have had the chance to grow over time. Our theory is in line with the findings of Wentzensen et al.,[31] who observed a difference in lesion size between women with HPV16 and other types in CIN2, but not in CIN3. Zaal et al. [22] found that with conventional visual colposcopy, no difference in sensitivity for CIN2+ lesions caused by different HPV types was found. However, the sensitivity of dynamic spectral imaging colposcopy for CIN2+ was higher in HPV16-positive than in non-16 hrHPV-positive women which may be explained by the better identification of small CIN2 lesions with this technique.[22] This is in agreement with our finding that in this population of women with abnormal cervical screening results, the visual colposcopic appearance of HPV16-positive women does not differ from that in women infected with other HPV types. Studies have shown that HPV16 is the genotype with the highest oncogenic potential because it is associated with the highest risk of the development of CIN3 and cancer and has the highest worldwide attribution for CIN2+ and ICC.[16, 19]

Conclusions

In this cross-sectional study in women with abnormal cytology and referred for colposcopy according to current European screening practice, we found that the sensitivity for detecting CIN2+ during colposcopy is similar in HPV16-positive and non-16 hrHPV-positive women. Irrespective of HPV status, all studied colposcopic characteristics were associated with CIN2+, in particular acetowhitening and lesion size. Furthermore, the mean age of HPV16-positive women with CIN2+ was significantly lower than in the non-16 hrHPV-positives. These findings confirm the important etiologic role of HPV16 in the development of cervical neoplasia, although HPV16-related CIN2+ lesions are not easier to detect during colposcopy. This might be a result of more rapid development of HPV16-related CIN2+. However, the performance of colposcopy in detecting these lesions is not improved, which may limit the benefit of this knowledge during colposcopic examination. As there is debate as to the accuracy of colposcopy in women vaccinated against HPV16 and 18, this study shows that colposcopic performance is similar for any hrHPV-type-positive woman.[35] To improve our understanding of the development of CIN2+ and CIN3+ related to HPV16 and other carcinogenic types, we are currently performing lesion-based genotyping using LCM-PCR.

Acknowledgements

We thank Agata Rodriquez, Nadia Abu-Lhiga, Hospital Clínic; Frank Smedts, Reinier de Graaf Groep; Edita Pirog, Weill Medical College of Cornell University, for their clinical assistance. We thank Lorena Marimon, Hospital Clínic; Hanna Breijer, Leids Cytologisch Pathologisch Laboratorium; Daan Geraets, Peter Lanser, DDL Diagnostic Laboratory; Karin van Kampen; Reinier de Graaf Groep, for their technical assistance. We wish to thank all colposcopists, laboratory personnel and the women who participated in this study.

Disclosure of interest

The authors declare no conflict of interest.

Contribution to authorship

J.M. and R.B. were responsible for drafting the manuscript with the assistance of J.B. for statistical analysis. A.T., M.P., I.A., B.H., R.B. and J.M. performed colposcopies. J.O., M.vdS., J.L. and D.J. were responsible for histological reviewing of all biopsies. W.Q., B.H., I.A., J.O., A.T., T.H., R.V., N.W. and M.S. designed the study. All authors critically reviewed the manuscript.

Details of ethics approval

Medical ethical approval from the institutional ethics boards of Hospital Clinic, Barcelona, Spain, and from Reinier de Graaf Groep, Voorburg, The Netherlands, was received (protocol number 10-023). The study was registered in the Dutch trial registry (NTR3464).

Funding

This research was funded by the Stichting Pathologie Ontwikkeling en Onderzoek (SPOO) Foundation, The Netherlands. The sponsor had no involvement in study design, data interpretation, writing of the manuscript, or decision to submit the manuscript for publication.

Ancillary