Editorials

Cervical sampling devices

In 1994-5, 4.5 million cervical smears were examined in England; over 350 000 (7.9%) were deemed inadequate.¹ Inadequacy rates reported by the 183 laboratories ranged from 0.2-35.5%. Such variation is unacceptable and must in part reflect different reporting criteria. Guidelines that should lead to a greater uniformity in reporting have since been circulated.² The rates also depend, however, on the quality of smear taking, and there is room for improvement here too.

In this week's BMJ, Buntinx and Brouwers (p 1285) review the relation between sampling devices and detection of dyskaryosis.³ The data suggest that extended tip spatulas (such as the Aylesbury) should be used in preference to Ayre spatulas and that brushes may be beneficial when used in conjunction with spatulas but that they should not be used alone. Here I will consider the appropriateness of combining results from studies with very different designs and the appropriateness of the endpoints used to evaluate screening.

The ideal sampling device would maximise the amount of cervical cancer prevented while minimising the costs of screening. No randomised study has evaluated prevention of cervical cancer directly; all have relied on surrogate endpoints. The best surrogate is perhaps the number of women treated for (histologically confirmed) high grade cervical intraepithelial neoplasia. Even this imperfect endpoint is not available in most studies; instead they rely on the rates of cytological abnormalities detected. A device associated with a higher rate of dyskaryosis would be judged to be superior, even if it were no better at cancer prevention, despite the costs (financial and psychosocial) of additional referrals. A good surrogate endpoint must be an accurate predictor of cancer prevention. Additionally, the chances of preventing cancer given the surrogate should be the same for all sampling devices in the study.⁴ Suppose one sampling device picked up additional cases of mild dyskaryosis based on cells sampled some distance from the transformation zone. If such cases were less frequently associated with progressive disease, the surrogate would be inappropriate.

Dey et al recently argued that inadequacy rates could be used as a surrogate for smear quality and that smear quality may be more appropriate than dyskaryosis for assessing cancer prevention.⁵ Although reducing the number of repeat smears would have clear cost benefits, one must be careful not to overinterpret the clinical importance of a reduced inadequacy rate. Mitchell and Medley showed that the incidence of cervical intraepithelial neoplasia in 20 000 women with a previous negative smear was not significantly different in those whose initial smear did or did not lack an endocervical component.⁶ It is now accepted that a report of "inadequate" should not depend solely on the presence or absence of endocervical cells, but it is still doubtful whether inadequacy rates can be considered a surrogate for screening efficacy.

The rates of dyskaryosis in Buntinx and Brouwers' paper range from under 1% in a screening setting to over 85% in a study of women referred with abnormal cytology.³ Combining relative risks from settings with such diverse underlying rates is hardly meaningful--a relative risk of 2 is impossible when the baseline is 80%. The use of odds ratios, while still problematic, seems more appropriate (table). The sampling device most suitable for routine screening may not be optimal for women who have been previously treated for cervical intraepithelial neoplasia. Data from the overview suggest that, whereas there is little advantage from using a brush in addition to a spatula in routine screening, the benefit in women referred with a previous abnormal smear may be more substantial (table).

Table--Odds ratio for various sampling devices for detecting mild dyskaryosis or worse relative to extended tip spatula alone. Odds ratios have been
calculated by pooling all studies and adjusting for underlying rate in each study.
---------------------------------------------------------------------------------------------------------------------------------------------------------
                                                All studies                         Screening only                             Referral only
---------------------------------------------------------------------------------------------------------------------------------------------------------
                                                         Odds ratio (95%
                                      No of smears        confidence           No of smears    Odds ratio (95%     No of smears        Odds ratio (95%
                                         taken             interval)            taken           confidence            taken             confidence)
---------------------------------------------------------------------------------------------------------------------------------------------------------
Extended tip                             27 939              1.00               11 302              1.00               905                  1.00
Ayre                                     14 329        0.87(0.79 to 0.96)         1232       0.49(0.24 to 1.01)       1039            0.92(0.69 to 1.23)
Ayre plus*                               11 459        1.09(0.97 to 1.23)         7373       0.57(0.26 to 1.29)        911            1.63(1.24 to 2.15)
Extended tip plus*                       12 023        1.08(0.95 to 1.22)         7342       1.14(0.77 to 1.69)        954            1.77(1.27 to 2.46)
Brush or swab                              2302        0.75(0.62 to 0.90)         1050       0.32(0.15 to 0.70)       1252            1.04(0.79 to 1.36)
Cervex                                   10 054        1.05(0.96 to 1.16)         3381       1.00(0.62 to 1.61)        280            1.37(0.87 to 2.14)
Cytopick                                   3406        1.08(0.68 to 1.73)         3406       1.10(0.66 to 1.82)          0           --
Bayne                                      4320        1.12(0.80 to 1.57)         4320       0.59(0.25 to 1.40)          0           --
---------------------------------------------------------------------------------------------------------------------------------------------------------
*Cytobrush or cotton swab.
Note that some studies had a mixture of screening referral smears or were conducted in gynaecology clinics and have only been included as "all studies."

Registrations of adenocarcinoma of the cervix have increased substantially in both Britain and the United States,⁷ and there is concern that cytological screening is less effective in preventing adenocarcinomas. Whereas the transformation zone must be adequately sampled for identification of precancerous squamous lesions, adenocarcinomas are likely to originate further up the endocervical canal. Thus there should be particular interest in the ability of sampling devices to pick up glandular lesions.

Testing for human papillomavirus is thought to be less reliant on adequate sampling of cervical cells, but as long as screening is based on cytology it is important for smears to be taken by trained practitioners using an appropriate device (such as an extended tip spatula), supplemented by a brush whenever the transformation zone is not visible. "The most important variable is probably the operator's skill."⁸ Screening programmes should monitor the inadequacy rates of smear takers, and anyone with a particularly high rate relative to that of the local laboratory should be offered retraining. Cervical screening in Britain has improved considerably since 1988, and it is probably preventing some 2000 cases of invasive cancer each year.⁹ Much can still be done to improve the quality of smears. It is hoped that Buntinx and Brouwers' paper will lead to the universal replacement of the Ayre spatula.

Research scientist Department of Mathematics, Statistics and Epidemiology, Imperial Cancer Research Fund, London WC2A 3PX

Peter Sasieni 

1. Government Statistical Service. Cervical screening programme, England: 1994-95. Department of Health, 1996. (Statistical bulletin 1996/3.)
2. Herbert A. Achievable standards, benchmarks for reporting, criteria for evaluating cervical cytopathology. Cytopathology 1995;6:(S2)7-32.
3. Buntinx F, Brouwers M. The relation between sampling device and detection of pathology in cervical smears. BMJ 1996;313:1285-90. [Abstract/Free Full Text]
4. Prentice R. Surrogate endpoints in clinical trials: definition and operating criteria. Stat Med 1989;8:431-40. [Medline]
5. Dey P, Collins S, Desai M, Woodman C. Adequacy of cervical cytology sampling with the Cervex brush and the Aylesbury spatula: a population based randomised controlled trial. BMJ 1996;313:721-3. [Abstract/Free Full Text]
6. Mitchell H, Medley G. Longitudinal study of women with negative cervical smears according to endocervical status. Lancet 1991;337:265-7. [Medline]
7. Beral V, Hermon C, Munoz N, Devesa SS. Trends in cancer incidence and mortality: cervical cancer. Cancer surveys 1994;19/20:265-85.
8. Wolfendale M. Cervical samplers: Most important variable is probably the operator's skill. BMJ 1991;302:1554-5.
9. Sasieni PD, Cuzick J, Lynch-Farmery E. Estimating the efficiency of screening by auditing smear histories of women with and without cervical cancer. Br J Cancer 1996;73:1001-5. [Medline]