Has Cytology Become Obsolete as a Primary Test in Screening for Cervical Cancer?

Rebolj and Lynge ¹ justifiably carefully examine the published results from the landmark randomized trial of cervical screening in rural India. ² They, like others, question why human papillomavirus (HPV) testing is associated with lower cervical cancer mortality when it is not associated with lower cervical cancer incidence or with higher rates of precancer (cervical intraepithelial neoplasia [CIN] grade 2 or 3) detection. Interestingly they calculate the proportion of unscreened women in the HPV arm and in the cytology arm diagnosed with cervical cancer. The non-significantly greater proportion in the cytology arm (RR 1.4, 95% CI 0.88–2.2) could, as the authors suggest, indicate that the cytology arm in this cluster randomized trial was at greater risk than the HPV arm. (They note that the proportions not screened in each arm are almost identical.) Alternatively, if the two arms were comparable, finding more cancer among non-attendees in the cytology arm might indicate that there should have been fewer cancers among attendees in the cytology arm.

Rebolj and Lynge's second possible explanation for the relatively poor performance of cytology is the observed poorer survival in the cytology arm. Although it is true that overall survival was worse in the cytology arm than the HPV arm, this can largely be attributed to the greater number of stage II or worse cancers in the cytology arm (n = 58) compared with the HPV arm (n = 39). Unfortunately, no details on the relative numbers of stage II, III and IV cancers are given.

While it is important to try to understand the lack of internal consistency between results in this trial, the findings about the relative benefits of cytology and HPV testing should not be taken out of context of the very large number of other comparative studies. The two most widely reported results are that compared with cytology, HPV testing is more sensitive but less specific at detecting high-grade CIN. For instance, Cuzick et al. ³ summarizing the results of trials in over 60,000 women, found the sensitivity of HPV testing to be 96% compared with 53% for cytology, while the specificity of HPV testing was 90.7% compared with 96.3% for cytology. Similar results have been obtained from more recent split sample trials, and the randomized comparisons of HPV testing with cytology also find better detection rates in the HPV arm but also higher ‘false-positive’ rates. Perhaps more telling is the variability in the sensitivity of cytology between studies compared with the consistently high sensitivity of HPV testing. At its best cytology can be very sensitive (77% in HART ⁴ ), but even within European studies the sensitivity was as low as 19%. ⁵ The problem is that cytology requires skill, concentration and judgement, and high quality is only achievable with a substantial investment in quality assurance. In one study in Peru, for instance, routine cytology read in a local laboratory had a sensitivity of 42.5% compared with 80.3% for liquid-based cytology read in a specialist cancer hospital laboratory in Lima. ⁶ It is noteworthy that the Indian trial put great emphasis on training and quality assurance,^2,7 which suggests that the cytology standards achieved in that trial would not be sustainable were such screening to be rolled out to tens of millions of women across India.

The reasons for the poor performance of cytology in many countries include poor sample collection, poor slide preparation and poor quality screening and review. Cytology relies on cells having been sampled from an area of high-grade CIN which may only cover a small proportion of the cervix: poor sample taking can mean that no abnormal cells are collected. By contrast HPV infection appears to colonize the whole cervix. Brink et al. ⁸ demonstrated the reasonable performance of HPV testing on self-collected vaginal samples contrasting the hopeless performance of cytology on such samples. Thus in the absence of existing infrastructure and a proven record of high-quality mass cytology for cervical screening, the case for cytology is limited. Taking into account the cost of quality assurance, it is not clear that cytology is any cheaper than HPV testing.

The issue of the low specificity of HPV testing needs addressing. One simple solution is to limit HPV testing to women aged over 30 or 35. In the European and North American overview, the specificity of HPV testing was 85.8% in women aged under 35, but 92.8% in women aged 35–49 and higher still in older women. ³ The standard approach to dealing with the low specificity of HPV testing is to run a second test on the screening sample in women who are HPV positive. Only those who are also positive on the triage test are referred to colposcopy (or immediate treatment), the others are offered repeat screening at a shorter interval (12 months say). The best evaluated triage test to date is cytology. In the Finnish randomized trial, the specificity of the HPV DNA test with cytology triage was equal to that of conventional cytology. ⁹ Reserving cytology for triage testing would require a smaller workforce and it would be easier to maintain high quality. Indeed the same trial found that in the HPV arm, where cytology is read only on HPV-positive samples, abnormal cytology lead to a higher rate of CIN2+ detection (80/20,065) than when all samples were screened by cytology (55/19,819). ¹⁰

HPV testing is not only more sensitive for high-grade CIN in cross-sectional studies. HPV-negative women remain at low risk of subsequent disease for much longer than cytology-negative women.^11,12 Thus a screening programme could safely extend the interval between screens if it were to switch from cytology to HPV testing. Such an approach would further improve the lifetime specificity of screening because even if the 30% more women are referred to colposcopy after each screen, with only half as many screens over a lifetime, the lifetime risk of referral to colposcopy would be only two-thirds of what it is currently.

Rebolj and Lynge are correct to point out the internal inconsistencies in the Indian trial results, but that does not justify their claim that ‘cytology will still be an attractive primary test for cervical screening’. Very recently, the Italian randomized controlled trial reported nine invasive cervical cancers during follow-up in the cytology arm compared with none in the HPV testing arm. ¹³ Cytology is not an attractive primary screening test because it is subjective, requires a high level of skill and performs extremely poorly in the absence of a huge investment in quality assurance. HPV testing by contrast can be performed on self-collected samples, processed in almost any laboratory and provides reproducible and objective results. ¹⁴