Human Papillomavirus testing for the management of low-grade cervical abnormalities in the UK—Influence of age and testing strategy

1. Introduction 

Appropriate management of low-grade cervical abnormalities has remained a contentious area in the field of cervical cancer screening. Although, the majority of such abnormalities will clear in time, a proportion will develop or harbour high-grade lesions that warrant investigation (Kinney et al., 1998, Rana et al., 2004).

Persistent infection with high-risk (HR) types of Human Papillomavirus (HPV) is necessary for the development of cervical cancer and its pre-cursor lesions; cervical intraepithelial neoplasia (CIN). The practical application of this being that HPV testing is very sensitive for the detection of such lesions. In the USA, highly influential prospective cohort studies, such as ALTS have resulted in the routine practice of performing HPV testing on women with a cytology result suggesting atypical cells of undetermined significance (ASCUS), with a positive HPV result constituting a basis for colposcopic referral (Kulasingam et al., 2006).

Although this evidence based management has been adopted in several centres within the US, significant changes to screening policy must be performed within national contexts. Time and cost saving policies from countries with opportunistic screening cannot be applied to countries where organised, successful call/recall screening is in place, such as the UK. In addition the relevance of study and policy comparison is diminished between countries that have different cytological disease classification systems and disease assessment strategies. For example, the ASCUS category is not used in the UK and approximates most closely, but not exactly with a result of borderline nuclear changes. Also, the provision of liquid based cytology (which has recently been rolled out across the UK) or conventional smear taking varies between countries and could alter the provision and effectiveness of HPV testing. In a recent editorial, Schiffman and Castle (2006) emphasised the notion that although the “technical efficacy” of HPV triage has been established, its cost effectiveness compared with alternative management strategies (involving cytology and colposcopy) is likely to vary between populations.

Kim et al. (2005) created a computer based model to assess how a cervical cancer screening system, which incorporated HPV testing, would compare with the existing screening strategies in four European countries, including the UK. Using national data for each country, the authors found that HPV triage of equivocal cytology results could incur health benefits, in a cost effective way, in all four.

Very recently, the results of the NHS UK HPV pilot studies were reported, which were designed to assess the effect of HPV triage in women with a cytology result of borderline changes or mild dyskaryosis, collectively termed as low-grade disease (Legood et al., 2006; Liquid Based Cytology/Human Papillomavirus Cervical Pilot Studies Group, 2006). To an extent, the findings consolidate the data of Kim et al. (2005) and are cautiously optimistic in that HPV testing in this group is likely to be cost effective. However, the authors also predict an increase in lifetime colposcopic referral, the management of which should be considered seriously before any policy decisions are made.

This leads us to the question of whether HPV triage of borderline/mild smears should be offered to all who are of screening age, or whether it should be age limited to reduce overload at colposcopy. Certainly, it is generally accepted that young women (<30 years) experience high levels of HPV infection, the significant majority of which is considered to be transient (Baseman and Koutsky, 2005). Indeed, in a study of the routinely screened population in Edinburgh, we found high-risk HPV to be detectable in around 35% of women under 30 years who were cytologically normal (Cuschieri et al., 2004). A minimum age for HPV triage could be supported should there be evidence that HR-HPV positive low-grade abnormalities in young women are less likely to progress and/or harbour significant lesions.

There is also the question as to what HPV tests would be appropriate for routine clinical testing now and in the future. The hc2 test has been used for large randomised controlled trials, and is used widely in the US, yet, as the market expands and more tests become available, there is a need to assess their utility. Moreover, it is good laboratory diagnostic practice to have more than one test available for individual targets.

The remit of this study was therefore to compare the concordance of two CE marked HPV detection tests using 322 archived cytology specimens graded with borderline changes. The second phase was to analyse a subset of these specimens longitudinally, in order to correlate disease persistence and clearance with age and HPV status by the two detection tests.

2. Materials and methods 

3. Results 

3.1. Cross-sectional comparison of HPV positivity by the two detection tests 

Out of the 322 specimens analysed, slightly more were found to be HPV positive by application of the AMPLICOR test compared with the hc2 test with 235 (73%) and 219 (68%) testing positive, respectively, although using a McNemars test this fails to reach statistical significance at the 5% level (p=0.056). A two-way comparison of positivity in both of the tests is depicted in Table 1. The two testing methods agreed in 80.7% of cases, 95% CI (76.0, 84.9).

Table 1. Comparison of HPV positivity in specimens graded borderline as determined by the hc2 and AMPLICOR

	AMPLICOR result	hc2 result
		Negative	Positive	ALL
	Negative	64	23	87
	Positive	39	196	235
	All	103	219	322

3.3. Relation of HPV status to disease outcomes 

We found evidence of a significant association between an HPV positive result by either method and evidence of disease: p<0.001. Of the 190 women in the subset, 141 (74.2%) were HPV positive by one or both tests, as detailed in Table 2.

Table 2. Frequency and distribution of disease outcomes after an average of 3.5 years in HPV positive vs. HPV negative women who had a borderline report

		HPV negative	HPV positive	All
	Clearance of abnormality (Group 1)	43	80	123
	Evidence of persistent low-grade abnormality (Group 2)	4	29	33
	Evidence of high-grade abnormality (Group 3)	2	32	34
	All	49	141	190

In the HPV negative group 8.2% (4/49) showed evidence of persistent low-grade abnormalities compared with 20.6% (29/141) in the HPV positive group. The difference in proportions being 12.40% (95% CI of 2.2–22.6%), p=0.05, Fisher's exact test.

In terms of high-grade disease, 4.1% (2/49) of the HPV negative women developed high-grade abnormalities compared with 22.70% (32/141) in the HPV positive group. The difference in proportion being 18.6% (95% CI of 9.8–27.4%), this difference was statistically significant (p=0.002, Fisher's exact test).

3.4. Age in relation to disease outcomes 

Women were assigned to “follow-up” Groups 1–3 by age category (Fig. 1). A χ2 test was used to determine if there was evidence of an association between age group and clearance of their borderline abnormality, irrespective of HPV status. No evidence of association was found (p=0.68). Confining the analysis to women who were HPV positive (by hc2 test and/or AMPLICOR test), there is no evidence that age was a factor in determining those cases that went on to develop high grade disease (p=0.44). However, differences would need to be large for this study to have sufficient power to detect them.

View Large Image Download to PowerPoint

Fig. 1. Outcomes of women who received a borderline cytology report from whom subsequent pathological follow up data was available. Group 1 (clearance without intervention), Group 2 (persistent low-grade abnormality) and Group 3 (development of CIN2 or worse).

3.5. Sensitivity and specificity of the two HPV detection tests for the detection of future high-grade disease 

The AMPLICOR test identified more HPV positive samples than the hc2 test. We compared the tests with regard to their ability to predict which borderline cases would show high-grade disease 3 years later (Table 3).

Table 3. Sensitivity of two HPV detection techniques for the identification of high-grade disease within 3.5 years of being diagnosed with borderline changes

		AMPLICOR (%)	hc2 (%)
	Sensitivity	82.4	82.4
	Specificity	36.5	44.2

The specificity of both tests (as depicted in Table 3) are low. However, it is important to appreciate that these measurements relate to a “one-off” HPV test for the identification of high-grade disease (within 3.5 years of a borderline result) in a population of women with a mean age of 31 years. Overall the results suggest that the AMPLICOR and hc2 tests were equally sensitive, while the hc2 test was more specific for the detection of high-grade disease in the future. However, these differences did not reach significance.

4. Discussion 

We have shown that for detection of HR-HPV within cytology specimens with borderline changes, the AMPLICOR test detected slightly more positives than the hc2 test (73% versus 68%). However, this difference did not reach significance, as has been previously described (Sargent et al., 2004). The higher reported analytic sensitivity of the AMPLICOR test is higher than the hc2 test and could account for this observation. Both tests performed similarly with respect to the longitudinal identification or prediction of high-grade disease.

In terms of practical issues associated with implementing either of these tests within screening settings, both are amenable to high-throughput – an essential requirement – and the equipment necessary for their execution has similar footprints. There is a PCR step in the AMPLICOR test, so specific zoning of pre- and post-PCR work is advised. There is also a nucleic acid extraction step in the AMPLICOR test, compared to a (equivalent) sample denaturation step in the hc2. A potential advantage of a specific extraction is that a nucleic acid resource is created, which could potentially be used for other tests. High-throughput robotic systems for sample processing and/or extraction, which feed into the hc2 and the AMPLICOR tests, have been also developed.

To date, little has been published on use of the AMPLICOR test in clinical settings although in a paper by Monsonego et al. (2005), the authors assessed sensitivity and specificity of the AMPLICOR test for the detection of high-grade lesions in women referred to colposcopy due to a preceding abnormal smear (using biopsy and/or LEEP as gold standard). They concluded that the AMPLICOR was highly comparable to published evidence of hc2 performance.

Our study differs from that of Monsonego and colleagues in that the two tests were compared directly i.e. using the same sample set. Also, concurrent histology was not performed in our study at the time of identification of borderline changes, so we cannot determine whether high-grade lesions developed later or were present, but undetected at the time of the borderline report.

Using single-test or combined HPV positive data, we found that HPV positive women were significantly more likely to develop CIN2+ than HPV negative women. However HPV positivity in smears with borderline changes will highlight low-grade disease (either progressing or regressing) and asymptomatic infection as well as high-grade disease. While the number of HPV positive borderline cases was high (around 70%), the specificity of a single HPV test for the identification of CIN2+ at an average of 3.5 years from a borderline report was low, irrespective of which test was used. The question, therefore, is how can one improve specificity of HPV mediated triage of low-grade smears? Limiting triage to older women may not be the answer as we found that younger women with borderline changes were no less likely to develop CIN2 or worse than older women. Moreover, this finding supports the work of a larger American study, where reflex HPV testing in 2309 ASCUS specimens (taken from women <25 to >64), was found to detect the same proportion of CIN2–3 in young women as elderly women (Eltoum et al., 2005). It is possible that the CIN2+ detected in the younger women in both of these studies was more likely to regress naturally, but this will be difficult to ascertain due to the usual policy of treatment for any CIN2 or worse.

This study is clearly not designed or powered to define an age-bracket and/or testing schedule through which cost effective HPV triage of low-grade smears in the UK could be achieved. The results of the UK pilot studies are encouraging and we await the results of the randomised control trials: TOMBOLA and ARTISTIC, also designed to assess the utility of HPV triage in the UK shed to more light on these issues. Certainly our study has shown that if HPV testing is introduced in the UK commercially available, potentially high-throughput broad-spectrum HPV DNA tests exist—two of which (described in the article) would appear to show comparable sensitivity for the detection of future high-grade disease within a context of borderline abnormalities. However, to prevent excessive use of a readily available test, informed provision of HPV testing is required and the search for better markers of clinically significant infection should be continued.