Comparison of the performance of carcinogenic HPV typing of the Roche Linear Array and Qiagen LiquiChip® HPV assays

Background Cervical cancer is caused by high-risk types of human papillomavirus (HPV). DNA testing of such high-risk types of HPV could improve cervical screening.The aim of the study was to compare the sensitivities and positive predictive values of two commercially available typing assays (Qiagen LQ and Roche LA) and to comparatively assess the distribution of HPV types with these two assays. Methods The study population comprised 311 ASCUS + women with abnormal pap tests who were HCII positive and who were admitted to three European referral gynecology clinics between 2007 and 2010 (Madrid, Marseille and Milan). All patients underwent LQ and LA tests. Results The sensitivity of the two assays for HPV typing was 94% for LQ and 99% for LA (compared with HCII). The overall concordance between LQ and LA was 93%. The three prevalent genotypes, HPV16, HPV18, and HPV31, were identified with a high concordance using the two assays: kappa 0.93, 0.83, and 0.91, respectively. Mixed genotypes were more frequently detected by LA than by LQ: 52% vs. 18%, respectively (p < .0001). Conclusions These assays have a good clinical sensitivity for detecting HPV types in CIN2+ patients and allow the virus type to be detected in the same experiment. Our study revealed no significant difference between LQ and LA for CIN2+ or CIN3+ diagnosis, indicating similar distributions of HPV types and a mixed genotype detection that is higher for LA than for LQ.


Background
Cervical cancer is caused by high-risk types of human papillomavirus (HPV). DNA testing of such high-risk (HR) types of HPV could improve cervical screening [1][2][3]. It has been demonstrated that either HPV HR type 16 or type 18 has a high CIN2+ predictive value [4][5][6]. The US guidelines suggest that if an FDA-approved genotyping test were available, it would be reasonable to use it in women 30 years old and older with negative cytology results and positive HPV tests. The detection of HPV 16 and/or 18 in that setting would be an indication for referral to colposcopy [7]. In this context, the Linear Array HPV (LA) (Roche) Genotyping Test, which detects 37 HPV genotypes by reverse line blot (RLB) hybridization, correlates in performance to the HPV HCII test for detection of CIN2+ [8][9][10][11][12]. Tools based on different technologies (reverse dot blot: LA, Roche; DNA-chip: Papillocheck ® HPV-screening, Greiner; clinical arrays: CLART ® Genomica) serve to assess the distribution of the HPV genotypes [13,14]. A novel commercial test (Qiagen LQ ® ) for the identification of 18 HR-HPV types on GP5+/6 + −PCR products was developed and has been analytically compared to the established RLB genotyping assay [15]. Godinez et al. recently performed clinical validation of LQ in women >40 years old [16]. However, no clinical validation has been performed in women aged >18 years old with an abnormal pap test.The primary aim of our study was to compare the sensitivities and positive predictive values of two commercially available typing assays (LQ and LA) comparatively with HCII using CIN2+ as the clinical cut-off. The secondary aim was to comparatively assess the distribution of HPV types with these two assays.

Patients
The study population comprised 311 ASCUS + women who were admitted to three European referral gynecology clinics between 2007 and 2010: 158 patients from Madrid (Spain), 58 from Marseille (France), and 95 from Milan (Italy). The criteria for eligibility were abnormal cervical smears, HCII-positive result, and referred for colposcopy with histology. A colposcopy, histology, and HCII test were performed independently at each study center (i.e., Madrid, Marseille, and Milan). Before colposcopy, a cervical sample was obtained using the ThinPrep method (Cytyc France Sarl, Roissy, France). The cervical scrapes were collected with the PreservCyt transport medium (Cytyc Corp., Marlborough, MA) and sent to a reference laboratory (Laboratoire Alphabio, Marseille, FRANCE), where the LQ and LA HPV tests were performed. All these tests were performed on the samples collected in PreservCyt liquid media for liquid-based cytology (ThinPrep). Informed consent was obtained from each participant according to ethics committee guidelines. This study was approved by the CPP Sud-Med I (Comité de Protection des Personnes Sud Méditerranée I) under the reference number 07 22.
In-house real-time PCR high-risk HPV genotyping An in-house Real Time (RT)-PCR genotyping method was performed on LA and LQ discordant cases. All PCR reactions were performed in a 20-μL volume using an ABI PRISM 7000 (Applied Biosystems). Each individual reaction contained 10 μL of Platinum ® Quantitative PCR SuperMix-UDG with ROX (Invitrogen, ref 11743), 5 μL of 0.4 μM primers + probes mix (MWG) and up to 5 μL of HPV DNA or sterile DNAse/RNAse-free grade water in non-template controls. The amplification profile was initiated by a 2-minute incubation at 50°C, followed by a 15-minute incubation at 95°C, and a two-step amplification of 15 seconds at 95°C and 60 seconds at 60°C for 45 cycles. For each 96-well PCR microplate, six patients could be tested for the 12 HPV HR (one HPV/well), the HBB gene, and a negative control. To optimize the reproducibility of the test, filling of the plates was conducted using a TECAN Freedom EVO 75 robot. Data were collected during the amplification step using ABI Prism 7000 SDS Software (Applied Biosystem).
In-house real-time PCR detects 12 HR-HPV genotypes: 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, and 59. To validate the performance of our in-house qualitative HPV type detection method using real-time PCR, standard curves were obtained by the amplification of a dilution series of ten million to ten copies obtained from CaSki cell lines for HPV 16, HPV DNA called HPV 475 (sample previously analyzed, infected with HPV18) and DNA extracted from PBMC as a positive control for HBB.
Sensitivities, specificities, positive and negative predictive values are reported in Table 2. Both tests displayed good sensitivity (LQ 94% and LA 99%). None of the comparisons of sensitivities and NPVs displayed statistically relevant differences between tests.
Only HPV genotypes common to LQ and LA were considered in the following analysis. Thus, HR-HPV detection was 95% for LQ and 99% for LA. Mixed HPV genotypes were more frequently detected by LA: 52% vs. 18%, respectively, p < .0001. Figures 1A and 1B show the absolute risk of CIN2+ or CIN3+ according to the presence of HPV16, HPV18, HPV31, or other HR-HPV types. HR-HPV types 16, 18, and 31 were chosen because of their high level of oncogenicity and for their high prevalence. The absolute risk of CIN2+ when typed with HPV16 by LQ was 60%, which is similar to the LA risk (61%). When typed with HPV18, the absolute risk of CIN2+ was 28% for both LQ and LA ( Figure 1A). Similarly, the absolute risk of CIN3+ when typed HPV16 by LQ was 23%, which is similar to the LA risk (25%). When typed HPV18, the absolute risk of CIN3+ was 6% for both LQ and LA ( Figure 1B).
The two tests displayed a high overall concordance rate of 93%. Among the 311 samples, 286 (92%) were HR-HPV positive by both tests, 3 (1%) were HR-HPV negative by both tests, and 22 (7%) were discordant. Among the 21 discordant cases HR-HPV negative by LQ, and seven were CIN2+ based on biopsy. The case that was HR-HPV negative by LA was CIN1 in biopsy. The in-house RT-PCR test was performed on LA and LQ discordant cases (Table 3). Among the 22 discordant cases, 21 were HR-HPV positive by LA, and one was HR-HPV positive by LQ. In-house RT-PCR was in accordance with LA in nine cases and in accordance with LQ in 13 cases. Table 4 shows the concordance between LQ and LA according to HPV genotype. The two tests are concordant for the majority of HPV genotypes except for HPV35 (Kappa = 0.41, p = 0.002), HPV53 (Kappa = 0.36, p < .001), HPV59 (Kappa = 0.29, p = 0.02), HPV73 (Kappa = 0.30, p = 0.04), and mixed genotypes (Kappa = −0.08, p < .001). LA detected HPV35 in 11 cases, but the Qiagen assay did not detect HPV35. Among these 11 cases, only two cases were HPV negative by LQ. In the nine remaining cases, LQ detected other HR HPV types. In the same manner, LA detected HPV53 in 35 cases and LQ in eight cases. All eight cases detected HPV53 by LQ were also detected to have HPV53 by LA. LA detected HPV59 in 18 cases. LQ detected HPV59 in three cases, and all were detected by LA. Among the 18 cases with detected HPV59 by LA, five were HPV negative by LQ. HPV73 was the last HPV   type for which discrepancies between LA and LQ were significant. LA detected HPV73 in 10 cases and LQ in three cases. Among the three LQ HPV73 cases, one was not detected as HPV73 by LA.

Discussion
In this study, an LQ HPV genotyping test was compared to the widely used LA test. LQ and LA were similar in HR-HPV detection (92% of concordant HR-HPV positive results). Although all samples were HCII positive, HPV negative results using the LA assay were found in 4 (1%) cases, compared with 24 (8%) for the LQ assay. Among the 311 samples, 286 (92%) were HR-HPV positive by both assays, 3 (1%) were HR-HPV negative by both assays, and 22 (7%) were discordant. Among the 22 LA and LQ discordant cases, in-house RT-PCR was in accordance with LA in nine cases and with LQ in 13 cases. A reason for the discordance between LA and LQ might be that among the 21 HPV-positive cases detected by LA but not by LQ, there were nine cases in which LA detected HR-HPV types LQ was not designed to detect. The LA and LQ tests were concordant in detecting the majority of HR-HPV genotypes except HPV35, HPV53, HPV59, and HPV73. The three prevalent genotypes, HPV16, HPV18, and HPV31, were found with a high concordance between LA and LQ: kappa 0.93, 0.83, and 0.91, respectively. These findings corroborate those reported by Godinez et al. [16]. All the cases where the listed HPV types were detected by LQ (35, 53, 59, and 73) were also detected by LA, except for one case where LQ detected HPV73 but not LA. Godinez et al. reported overestimation of LA for HPV52 and HPV53 because of cross-reactivity with other probes. We think it might explain the discrepancies between LA and LQ in detecting HPV35, HPV53, HPV59, and HPV73 [16]. Mixed genotypes were more frequently detected by LA than LQ: 52% vs. 18%, respectively (p < .0001). As previously reported, LA is known to have a higher power in identifying mixed genotypes, but the clinical implication of this higher power is not well established [8,9,16]. One hypothesis could be that the genotyping results are due to the nature of the genotype detection used by the test: LA uses a strip that is known to be more sensitive to detect a small quantity of DNA, instead of the liquid used with the LQ.
The analysis of absolute risk of CIN2+ and CIN3 revealed no difference between LA and LQ. The two assays are very similar, regardless of HPV type. As an example, the  absolute risk of CIN2+ when HPV16 positive (60%) is approximately two-fold the absolute risk of CIN2+ when HPV18 positive (28%). The overall agreements between LA and LQ genotyping assays indicated highly similar outcomes and were close to those in recent reports using different genotyping methods [8,9,[20][21][22][23]. However, the HR-HPV positivity rate was higher (not statistically) with the LA assay than with the LQ.
Genotype-specific results obtained with LA and LQ were highly similar, with 91% of concordant or compatible results.
Regarding the laboratory technical characteristics, LQ® uses multiplex, bead-based xMAP technology and an automated, high-throughput read-out with either the LiquiChip 200 workstation (QIAGEN, Hilden, Germany) or Luminex 100 IS System (Luminex Corporation, Austin, TX). The test was developed for identification of the 18 HR-HPV genotypes associated with cervical cancer by using GP5+/6 + −PCR products. A high level of agreement has been reported between the LQ test and the LA assay for detection and genotyping of 18 HR-HPV types in HCII positive specimens [15].
A limitation of this study might be its relative low number of patients included despite the specific inclusion criteria.

Conclusions
In conclusion, LQ and LA assays both have good clinical sensitivity for detecting HPV types in CIN2+ adult women. This study revealed no significant difference between LQ and LA for CIN2+ or CIN3+ diagnosis, in addition to indicating similar distributions of HPV types. These assays detect and determine the type of virus in the same experiment. The availability of HPV genotyping methods with demonstrated reliable clinical performance is a necessary prerequisite for HPV genotyping assays.