Although case–control studies have been extensively used for evaluating screening for cervical cancer, this approach has been criticised on the grounds that a bias in favour of screening could occur (Sasieni, 2001
, pp 93–105; Zappa and Ciatto, 2001
, pp 99–118). In the present study, nevertheless, the main aim was not to compare the protection in screened as compared to unscreened woman: on the contrary, we compared the relative protective effect of screening (a) for different histological type of cancer (squamous and adenocarcinomas) or (b) in different age groups, and we suppose that any potential bias should similarly affect different compared strata. In other words, we cannot exclude that our case–control approach may overestimate the effect of screening, but we do not expect such an overestimate to be different when comparing adenocarcinomas to squamous cancers or younger to older women. However, other differential biases are possible. If adenocarcinoma is more likely to be screen detected, this would bias results in favour of a preventive effect on adenocarcinoma (cases will have a full screening interval, whereas controls will on average be halfway between two screens); if high-risk young women are more likely to attend whereas high-risk older women are more likely to refuse screening, then such a difference could explain the result we have observed.
The present study shows a lower protection from cytological screening for adenocarcinoma than squamous carcinoma. The result is consistent across age groups, and confirms the findings of other case–control studies (Mitchell et al, 1995
; Sato et al, 1997
). From a public health point of view, this might be of minor concern as adenocarcinomas currently account for less than 25% of cervical cancers, but such a lower protective effect of screening should be kept in mind, considering that the incidence of adenocarcinoma is increasing in younger women. The investigation of the possible causes of a lower protective effect of cytological screening from adenocarcinoma (e.g. lower sensitivity of cytology, different sensitivity of smear sampling techniques, differences in sojourn time intrinsic of adenocarcinoma) is essential to support further action.
As far as the association of age with screening efficacy is concerned, we found that the relative protection of screening from invasive squamous carcinoma is shorter in younger (<40 years) than in older (
40 years) women. This finding confirms the results of a recent large study carried out in UK (Sasieni et al, 2003
), which recommended a reduction from 5 to 3 years in the rescreening interval for women below 50 years of age. In Italy, a 3-year rescreening interval is recommended at any age, and such an interval seems to grant sufficient protection. In the case of older age groups, our findings suggest that a longer interval (5 years) could be safely adopted.
The scenario of cervical cancer prevention is open to marked changes, due to the widely recognised aetiological role of human papilloma virus (HPV) and due to the possibility of the future use of HPV vaccination as a radical alternative to screening; but the role of HPV testing as a possible screening tool should not be dismissed. Attempts to improve screening sensitivity and efficacy in specific circumstances (younger age, adenocarcinoma) should not be based only on a more aggressive cytological approach (e.g. reduced rescreening interval, lower threshold for cytological abnormality prompting diagnostic assessment) but also consider the opportunity for adding HPV testing to Pap smear as a screening tool.