Human papillomavirus testing

Editor—In This Week in the BMJ of 14 April our study of human papillomavirus testing and the management of women with mildly abnormal cervical smears is described as an NHS pilot¹; it was not. In the accompanying editorial Manos expressed surprise that we advised caution in the clinical use of this test despite evidence for its role in managing women with borderline cervical smears.² We intended the context of our advice to be that of our paper: the management of women with borderline or mildly dyskaryotic smears.

As we stated, although there is evidence from the United States that human papillomavirus testing is useful in triaging the equivalent of borderline change, such testing has limited potential in triaging the equivalent of mild dyskaryosis (because of high positivity for papillomavirus (83%)).³ Caution would seem reasonable if these two types of mild abnormality are being considered together (as in the NHS pilot scheme based in England) rather than separately (as in the United States). We were disappointed that Manos's editorial said little about mild dyskaryosis.

We found a prevalence of papillomavirus (41%) and a test sensitivity (86%) for those with only borderline smears that were comparable to those of the Kaiser Permanente study mentioned by Manos, but we found a lower sensitivity (92%) and negative predictive value (89%) for younger women (29 subjects) than it reported. We confirmed high positivity for papillomavirus (75%) among those with only mild dyskaryosis, with a low positive predictive value (39%); the highest negative predictive value in this subgroup was 92% among 68 younger women (1 pg/ml cut off point).

We agree that our study design contributed to the prevalence of high grade disease (35%), but this figure is consistent with that in previous reports.⁴ Our subjects had had persistent mildly abnormal smears before being referred for colposcopy (reflecting United Kingdom guidelines). Although high grade lesions associated with such smears tend to be small (emphasising the need for accurate diagnosis),⁴ no evidence exists that persistence in itself selects lesions that are difficult to sample or identify. We used large loop excision of the transformation zone because it is more accurate than colposcopic biopsy (and yields more high grade disease) and allows treatment at the first visit.⁵ Each specimen was examined by at least two pathologists (with initial reporting by at least one and then review by GR).

We believe that human papillomavirus testing may prove useful in cervical screening in certain defined contexts, which may involve different methodologies including the polymerase chain reaction. We await the results of the NHS pilot scheme.