Comparison of the ScreenFire and Xpert HPV assays for the detection of human papillomavirus and cervical precancer among women living with HIV in Malawi

Cervical cancer incidence and mortality data demonstrate a dire global health inequity. Despite cervical cancer being highly preventable, in 2020, low-and middle-income countries (LMICs) accounted for 85% of the estimated 570,000 incident cases and 90% of deaths globally [1]. The burden of cervical cancer is particularly pronounced in sub-Saharan Africa (SSA), where HIV is endemic [2]. Women living with HIV (WLWH) have a higher incidence and persistence of high-risk human papillomavirus (HPV), the causative agent of cervical cancer [2].

The WHO recommends HPV testing as a primary cervical cancer screening method globally, including in LMICs [3]. Studies on the use of HPV-based screening in LMICs have demonstrated that HPV testing on self-collected specimens is comparable to provider-collected specimens for detection of cervical precancer/cancer [4, 5], is cost-effective [6, 7], and highly acceptable across many countries [8‐12].

Use of HPV testing for primary screening in LMICs is limited, however, with less than 5 African countries recommending HPV testing as a primary screening method [13]. This limited use is partly due to the cost and logistical challenges of incorporating HPV testing within existing “screen & treat” programs in LMICs [14, 15]. To increase the feasibility of HPV-based cervical cancer screening in LMICs, affordable, easy-to-use, and point-of-care HPV assays are needed in both facility and community-based screening programs. In addition, the assays ideally need to be able to perform HPV risk stratification through genotyping to inform triaging or management of HPV-positive individuals, based on the carcinogenic risk of each of the high-risk HPV types [16, 17]. This stratification is particularly important in LMICs where the “screen & treat” strategy is used and same-day treatment decisions are often necessary following primary screening to reduce patient loss-to-follow-up [18]. While several HPV testing technologies with full, extended, or partial genotyping capacity are available, including Aptima® (Hologic Inc, USA), Onclarity (BD Corporation, USA), and Cobas® 4800 (Roche Molecular Diagnostics, Switzerland) [19, 20], the Xpert HPV Assay (Cepheid, Sunnyvale, CA, USA) [21‐23], is one of two WHO-prequalified [24] HPV tests available for use in field conditions in LMICs. Advantages of the Xpert HPV assay include its ability to detect 13 high-risk HPV types (and HPV66, which is not high risk) in five risk-based channels, which enables risk stratification from a single cartridge in approximately 60 min. [25] The cartridge, which is prefilled with primers and reagents necessary for extraction, amplification, and detection of HPV DNA regions, significantly increases feasibility for use in remote settings without laboratory expertise. Additionally, Xpert® HPV testing is performed on Cepheid GeneXpert platforms, which are widely used for molecular diagnosis of tuberculosis in many LMICs, allowing the integration of both tests on the already available platforms. However, key limitations of the Xpert HPV include the price of the testing (approximately $15/assay [26], which is out of reach for most LMICs), the need for a separate PreservCyt medium for specimen preparation prior to testing (which adds logistical challenges and cost), and its inability to simultaneously test a large number of specimens. These limitations highlight a need for alternative assays for use in LMICs.

A promising alternative for rapid, low-cost, and high-volume HPV testing in LMICs is the use of isothermal amplification rather than traditional PCR technology [27, 28]. Isothermal amplification-based HPV assays have been developed by Atila Biosystems (Mountain View, CA, USA) and can potentially be more affordable than most currently available HPV assays [20]. There are two marketed formats: (i) multiplex detection of 15 high-risk HPV types with separate detection of HPV 16/18 in a single tube (AmpFire HPV assay), and (ii) the Ampfire HPV Genotyping assay with individual genotyping of 15 high-risk HPV types in four tubes [29]. The assays are Conformite Europeenne (CE) marked [30] and have been evaluated in analytical [31] and clinical studies [20, 27, 30, 32‐35]. These isothermal-based assays have been validated for dry specimen collection [36], and allow for storage of dry swabs at room temperature for two weeks [28], significantly increasing feasibility for use in LMIC settings.

The existing 15-type isothermal assay (AmpFire HPV) was recently redesigned for public health use as a 13-type assay (ScreenFire) with four channels based on differential cancer risk: (i) HPV 16, (ii) HPV 18/45, (iii) HPV 31/33/35/52/58, (iv) HPV 39/51/56/59/68 [28, 37]. This redesigned assay, performed in a single tube, was compared to an AmpliTaq Gold MY09-MY11 PCR-based HPV test on 453 provider-collected samples from Nigeria with very high agreement – the weighted kappa for ScreenFire versus AmpliTaq Gold was 0.90 (95% CI 0.86–0.93) [28]. Recently, ScreenFire was compared to Linear Array and TypeSeq using 2,076 provider-collected samples from the USA, demonstrating excellent clinical performance, with a CIN3 + sensitivity of 94.7% (95% CI 92.6–96.4) for ScreenFire, 92.3% (95% CI 89.7–94.3) for Linear Array and 96.0% (95% CI 93.9–97.6) for TypeSeq [37]. Similarly, ScreenFire was compared to Ampfire, Cobas and SeqHPV genotyping using Chinese samples to satisfy Meijer’s Criteria for clinical endpoint validation [38], and simulating the VALGENT framework for inter and intra-laboratory validation [39]. In this study using samples from the Chinese Multi-Site Screening Trial (CHIMUST), ScreenFire demonstrated excellent genotype-specific concordance when evaluated for clinical guidance in hierarchical fashion with both the Cobas 4800 and SeqHPV for < CIN2, CIN2, and ≥ CIN3 [40].

However, there is currently no data available on the performance of the ScreenFire HPV assay compared to Xpert among WLWH in real-world screening specimens. To inform the use of the ScreenFire assay for primary HPV testing in high HIV-burden settings, we compared the performance of ScreenFire with the Xpert assay using stored cervical samples for detection of HPV and cervical precancer among WLWH in Malawi.

Table 1 shows the overall and within-channel agreement between Xpert and ScreenFire HPV tests on the 315 stored provider-collected specimens. Of these 315 specimens, 221 (70%) and 236 (75%) were high-risk HPV-positive on Xpert and ScreenFire, respectively, representing 80% agreement on positives, with an unweighted Kappa of 0.59 (95% CI 0.50–0.69). When comparing agreement within the first three channels, which have identical HPV type groupings, percent agreement was highest (72%) for Channel 3 (HPV 31/33/52/58/35), with an unweighted kappa of 0.73 (95% CI 0.65–0.80), and lowest (60%) for Channel 2 (HPV 18/45), with an unweighted kappa of 0.71 (95% CI 0.60–0.83).

Table 2 demonstrates the comparison of overall and within-channel agreement between Xpert and ScreenFire on the 279 paired self-collected specimens that were Xpert HPV-positive at baseline. Of these, 252 (90%) were HPV-positive on ScreenFire. Of the 59 (21%) HPV16-positive on Xpert, 49 (18%) were HPV16-positive on ScreenFire. Concordance between the two assays in the HPV 18/45 and HPV 31/33/52/58/35 channels was relatively lower. Of the 49 (18%) HPV 18/45 positive on Xpert, only 38 (14%) were also HPV 18/45 positive on ScreenFire, and of 162 (58%) positive for HPV 31/33/52/58/35 on Xpert, 141 (51%) were positive on ScreenFire. The other channels could not be directly compared as they have different HPV types. A similar comparison was done, stratified by histologic diagnosis (Supplemental Table S2). Because Xpert HPV-negative self-collected specimens at baseline were not stored, percent agreement and kappa could not be calculated for Table 2.

Tables 3 and 4 show results of ScreenFire hierarchical HPV positivity by channel, stratified by the histologic diagnosis, using Xpert as the reference, for provider- and self-collected specimens, respectively. In these hierarchical analyses, the highest priority cases were CIN2+ with HPV 16. For the hierarchical analyses of provider-collected specimens (Table 3), among HPV 16-positives by Xpert, ScreenFire detected 12 of 15 (80%) with CIN2+ and 22 of 30 (73.3%) with < CIN2. Here, ScreenFire missed no HPV16 CIN2+ cases identified on Xpert (i.e., 3 HPV16 Xpert positive CIN2+ cases were negative on ScreenFire HPV16 but positive on another ScreenFire channel). Among 30 Xpert HPV16 positive, < CIN2 cases, 22 (73.3%) were also HPV16 positive on ScreenFire, 7 (23.3%) were positive for other channels on ScreenFire, and only 1 (3.35) was HPV-negative on ScreenFire (Table 3). Among the provider-collected HPV 18/45-positives on Xpert, 7 of 7 (100%) with CIN2+ were also HPV 18/45 positive on ScreenFire, and 21 of 26 (80.8%) with < CIN2 were also HPV 18/45 positive on ScreenFire (Table 3). For the HPV 31/33/35/52/58 channel, among 20 positives on Xpert with CIN2+, ScreenFire detected 19 (95%) as positive for the same channel, with only 1 (5%) testing HPV-negative on ScreenFire.

Among the 49 CIN2+ cases in our analysis, all of whom were HPV-positive on self-collected specimens at study entry (Table 4), 3 (6.1%) were HPV-negative on Xpert provider-collected specimens (Table 3). Of these 3 CIN2+ missed by Xpert, 1 was also ScreenFire HPV-negative on the self-collected specimen, while 2 were additionally picked up by ScreenFire, with a positive result on Channels HPV 16 and 31/33/35/52/58, respectively. As expected, there was more variation in channel positivity between Xpert and ScreenFire among women with < CIN2, who are known to have lower HPV viral loads than those with CIN2+ for HPV 16 and the HPV 16-related types. [45]

For the hierarchical comparison of ScreenFire and Xpert (reference standard) for self-collected specimens (Table 4), similarly high levels of agreement are demonstrated. Of 15 CIN2+ cases that were HPV16-positive on Xpert, all 15 (100%) were also HPV16-positive ScreenFire. Among 7 CIN2+ cases that were HPV 18/45-positive on Xpert, 6 (85.7%) were also HPV18/45-positive on ScreenFire, and 1 (14.2%) was positive on ScreenFire HPV 31/33/35/52/58 channel. For 23 CIN2+ cases that were HPV 31/33/35/52/58-positive on Xpert, 19 (82.6%) were positive for the same channel on ScreenFire, 3 (13.0%) were positive on ScreenFire HPV 39/51/56/59/68 channel, and 1 (4.3%) was missed on ScreenFire (tested negative).

In summary, all 49 CIN2+ cases were HPV-positive on Xpert on self-collected samples, while 2 (4.1%) were missed by ScreenFire (Table 4). Agreement on self-collected specimens between the two assays for the < CIN2 cases was also high for the 3 comparable channels (Table 4). Among the approximately 10% (N = 36) participants who were Xpert HPV-negative on self-collection at study entry and had a provider-collected specimen tested on Xpert and ScreenFire, all 36 were HPV-negative on Xpert provider-collected assay and all had < CIN2 on pathology. None of these Xpert HPV-negative specimens were available for ScreenFire testing. Hence, comparison with ScreenFire could not be done.

Table 5 compares the hierarchical ScreenFire HPV positivity for self-collected specimens to Xpert provider-collected specimens (reference), similarly stratified by histologic diagnosis. Similar to Table 3, ScreenFire missed no HPV 16 CIN2+ cases that were identified on Xpert. Out of 15 cases of CIN2+ that tested positive for HPV 16 on Xpert, 13 (86.6%) also tested positive for HPV 16 on ScreenFire. The remaining two cases tested positive on different ScreenFire channels: one for HPV 18/45 and the other for HPV 39/51/56/59/68. Among 7 Xpert HPV 18/45-positive CIN2+ cases, all were positive on ScreenFire, with 5 (71.4%) also positive on the 18/45 channel.

In this hierarchical comparison of ScreenFire self-collected to Xpert provider-collected results, there was also high agreement between the two assays for the HPV 31/33/35/53/58 channels. Of the 20 CIN2+ that tested positive for HPV 31/33/35/52/58 on Xpert, 17 (85%) were positive for the same Channel on ScreenFire, and 2 (10%) positive on a different ScreenFire channel (HPV 39/51/56/59/68 and HPV 16). Only 1 (5%) of the CIN2+ cases positive on Xpert on this channel was missed (tested negative) on ScreenFire. Similar to Table 3, 3 CIN2+ cases were negative on Xpert provider-collected samples, despite being positive on Xpert self-collected specimens. These samples were all positive on ScreenFire on the self-collected specimens.

In this analysis of stored cervical samples of primarily Xpert HPV-positive self-collected specimens among WLWH in Malawi, the ScreenFire HPV assay had good agreement with the WHO-prequalified Xpert HPV assay. ScreenFire demonstrated 80% agreement on HPV positives when compared to Xpert using paired provider-collected specimens. The agreement between the two assays was excellent for the HPV16 channel (unweighted kappa = 0.75) and good for HPV 18/45 (unweighted kappa = 0.71) and HPV 31/33/52/58/35 (unweighted kappa = 0.73) on provider-collected specimens. Among paired self-collected specimens, 90% of samples that were Xpert positive were also ScreenFire positive. Channel-specific concordance was highest for HPV 16 and slightly lower for HPV 18/45. In hierarchical analysis, ScreenFire demonstrated high concordance with Xpert for detecting CIN2+ cases in all channels, missing no HPV 16 or HPV 18/45 CIN2+ case that was positive on Xpert, on both self- and provider-collected specimens. All HPV 16-positive or HPV 18/45-positive CIN2+ cases on Xpert were HPV positive on ScreenFire, although not always for the exact channel. This delineation is important given the importance of HPV 16 and HPV18 in cervical carcinogenesis. Of note, among the 49 CIN2+ cases in our sample cohort, all of which were Xpert HPV-positive on self-collection by study design, 3 were HPV-negative on Xpert provider-collected specimens, of which one was negative on ScreenFire provider-collected specimen. This may be a false negative or possibly sampling error. When comparing the results of ScreenFire self-collected specimens to Xpert provider-collected specimens, all 3 CIN2+ cases that were HPV-negative on Xpert provider-collected specimens were HPV-positive on ScreenFire self-collected specimens.

With the demonstrated comparison, the ScreenFire HPV assay has several advantages that can significantly improve access to primary HPV screening in LMICs where the burden of cervical cancer is highest. ScreenFire can allow testing of up to 96 samples at a time, increasing feasibility for use in community-based campaigns or high-volume clinics at a relatively low-cost (~ 6 USD per test) [27], which is less than half the price of Xpert HPV tests. Additionally, ScreenFire’s extended genotyping capability, which we demonstrate to have good concordance to Xpert HPV and histology, if supported by larger validation studies, can be utilized for risk stratification in “screen and treat” programs in LMICs.

Compared to some other HPV typing assays available, ScreenFire is relatively simpler to use, does not require DNA extraction, and is compatible with the collection and transportation of dry swabs to the laboratory. These characteristics significantly improve the feasibility of using ScreenFire for extended HPV genotyping in LMIC settings with basic laboratory infrastructure or within community-based and mobile clinics. Of note, while a centrifugation step is necessary when processing frozen samples for ScreenFire testing, as in our study, this step is not required with use of fresh samples, as would be the case with point-of-care testing. Limitations of ScreenFire include the need for hand pipetting for sample preparation, which requires basic pipetting skills and individual mixing of amplification reagents, both of which can be subject to contamination under field conditions, reducing accuracy or providing invalid results. Additionally, at ~ 6 USD per test, while more affordable than other assays, this price may still be out of reach for many LMICs for use for primary screening.

While the benefits and limitations of the ScreenFire assay have also been discussed in a recent large cross-sectional study [37], our data add two very important aspects specific to LMIC settings. First, this is the first time, to our knowledge, that the novel redesigned ScreenFire HPV assay has been validated on samples from WLWH, who are known to have generally a higher prevalence of multiple HPV infections [46]. This population is crucial considering the high prevalence of HIV in many LMICs that bear the greatest burden of cervical cancer, and hence could most benefit from high-performing and affordable HPV testing. Second, this is the first time that ScreenFire has been evaluated on self-collected cervicovaginal samples. The use of self-collection is key when implementing primary HPV testing in low-resource settings because of the potential of vastly increasing screening coverage through community or home-based self-collection, when compared to provider-collected specimens, which require a pelvic examination in a clinical setting. Larger studies using ScreenFire on self-collected samples are ongoing, both on WLWH and HIV-negative women [47].

Among study limitations is that we compared the performance of the ScreenFire assay on samples which were previously frozen, which may have resulted in lower test positivity as compared to use of fresh samples. However, both assays are based on testing of DNA, which is highly stable, and storage was a tertiary-level laboratory with 24-h backup generators and a robust quality assurance program. Similarly, because the original study was not designed as a validation study, only Xpert HPV-positive and a random 10% of Xpert HPV-negative specimens on self-collection at baseline were stored and available for ScreenFire testing, possibly biasing our comparison as not all HPV-negative samples were stored for future testing. Lastly, ScreenFire testing in this study was performed at a tertiary-level laboratory in Malawi, with qualified personnel that may not represent field conditions in many LMICs. As such, it is possible that the use of ScreenFire under real-world conditions may yield lower agreement/accuracy compared to our results. To inform this and further guide the use of ScreenFire for HPV testing among WLWH in LMICs, a more extensive study is needed to evaluate test performance and implementation aspects, including the feasibility of testing across different LMICs settings and conditions.

In conclusion, the ScreenFire HPV assay performed well compared to a WHO-prequalified and clinically utilized HPV assay among WLWH in LMIC setting. If supported by larger validation studies, the ScreenFire HPV assay may be a promising option for increasing access to accurate and affordable, risk-based HPV screening and management through genotyping in LMICs.