When to start cervical screening: epidemiological evidence from Canada


Cervical screening is recognised as one of the most effective cancer screening methods available, with over 80% reduction in incidence and mortality when applied well.[1] However, guidelines in different countries vary in their recommendations about the age to start screening, the interval between screens, and when to stop, as well as details about when and how to investigate abnormalities.

The age to start screening is a vexed question internationally. In Canada, guidelines used to recommend screening from the beginning of sexual activity, or from age 18 years at the latest.[2] Australian guidelines still do, although revision is under consideration.[3] Other countries start screening at the age of 25 or 30 years.[4] A working group of the International Agency for Research on Cancer recommended screening should not start before age 25 years.[5] The Canadian Task Force on Preventive Health Care sought national evidence to assure women and clinicians that decisions about policy are appropriate for Canadian women.[6] This commentary describes this evidence.

There has been no randomised trial comparing Pap-test screening policies, and most research about cervical screening has concentrated on assessing its effectiveness between or within countries on an age-standardised basis. The effectiveness of screening by age must be evaluated to decide the age to commence screening, yet there is surprisingly little evidence that has been analysed in this way.

It is relatively easy to measure these effects in Canada. Cervical screening was adopted early in Canada, from young ages, with widespread and frequent testing, and good data were recorded. A laboratory was established in Vancouver in 1949: this was developed into a provincial programme, and the concept spread across the country, so that it was possible to demonstrate the effectiveness of screening relatively quickly.[7] Canada has better data on cancer than most countries: the recording of national mortality data has been in place since 1932, although this is subject to the vagaries and inaccuracies of death certification. Cancer registries were established early in various provinces, and complete national data have been available since 1972, around the time that screening became widespread.

Canadian women have been regularly screened: their doctors generally provided annual screening from the beginning of sexual activity. The Pap test was linked with the pill: women attended annually for a repeat prescription, Pap test, and pelvic examination. For those who did not obtain contraception, guidelines recommended screening during antenatal and postpartum care. Initially women had to pay, but since the establishment of National Health Care (Medicare) in 1986, medical examinations, including Pap tests, have been free of any charge to women, so there was no financial disincentive. However, when recommendations were made by national task forces to increase the interval between tests to 3 years,[8, 9] they were opposed jointly by the Canadian societies of obstetricians and gynaecologists, colposcopists, and gynaecological oncologists, until all the elements for organised screening were in place,[10] so the majority of doctors continued annual or biennial screening for women during their reproductive years and afterwards. High rates of screening were thus widespread,[11] but under-screened and unscreened women continued to account for half of the cervical cancer cases diagnosed, and were concentrated among those with various forms of social disadvantage: rural women, immigrants, and aboriginal populations[6]. Thus Canada demonstrates the effects of the maximum level of screening practically possible without organised registry systems, which are only now being developed in most provinces.

Initially, the incidence and mortality rates were similar to those currently found in third-world countries, so that 1.5% of women developed this cancer over their lifetime, and nearly 1% died from it, but mortality and incidence rates have dropped substantially since the institution of screening,[11] similar to countries that undertake less intensive screening.[1]

Cervical screening does not simply detect and lead to treatment of invasive cancer: pre-cancer changes are also detected, leading to their treatment, and thereby preventing the development of invasive cancer. However, many pre-cancerous lesions would regress, so many more are treated than would ever become invasive cancer.[12] Furthermore, the time of detection of early-stage cancer is brought forward, so that when screening is introduced, apparent incidence rises. Data from British Columbia extending back to 1970 shows an initial rise in incidence, to a peak during the late 70s and early 80s that was proportionally greater for younger women.[11] This timing corresponds with the introduction of colposcopy, when many small punch biopsies were taken and pathologists struggled with interpretation, until they became familiar with these new specimens, and standards for diagnosis were developed.

To answer the age-specific questions, we analysed the data by 5–year age groups, using year, time period, and birth cohort approaches. During the screening era the age distribution of cancer has changed, with substantial reductions over time, especially for women over the age of 30 years, where there has been a steady and successive drop in both incidence and mortality over time, and for each cohort, as shown in Figures 1 and 2.[11] Tables 1 and 2 show mortality and incidence by age for young women up to the age of 35 years, and the changes over the past 20 years.[11] Whereas the relative risk has reduced by 30% to 50%, the absolute risk reductions are very small. For women aged 20–24 years the mortality reduction is around one per million, and for women aged 25–29 years the mortality reduction is one per 100 000. Incidence reductions appear slightly greater, but are likely to have been biased by starting at the peak of the artifact noted previously. These correspond well with the findings from the surveillance, epidemiology, and end results (SEER) registry data in the USA,[13] and from using case–control methods in the UK.[14]

Table 1. Cervical cancer mortality changes in Canada since peak death rates (all rates per 100 000)
Age group15–19 years20–24 years25–29 years30–34 years
Deaths nMortalityDeaths nMortalityDeaths nMortalityDeaths nMortality
  1. a

    Number not released because the cell size was less than five.

  2. Reproduced in part from: Dickinson et al.[11]

1972–1976 a a 50.1300.6661.8
Relative rate reduction 0 0.33 0.65 0.74
Absolute rate reduction 0 0.1 1.1 3.5
Table 2. Cervical cancer incidence changes in Canada since national data have been available (all rates per 100 000)
Age group15–19 years20–24 years25–29 years30–34 years
New cases nIncidence rateNew cases nIncidence rateNew cases nIncidence rateNew cases nIncidence rate
  1. Reproduced in part from: Dickinson et al.[11]

Relative rate reduction 0.35 0.52 0.26 0.26
Absolute rate reduction 0.1 1.4 2.4 4.4
Figure 1.

Age-specific mortality from invasive cervical cancer in Canada, 1972–2006. Reproduced from: Dickinson, JA., Stankiewicz, A., Popadiuk, C., Pogany, L., Onysko, J., Miller, AB., Reduced cervical cancer incidence and mortality in Canada: national data from 1932 to 2006 BMC Public Health 2012, 12:992.

Figure 2.

Age-specific incidence of invasive cervical cancer in Canada, 1972–2006. Reproduced from: Dickinson et al.[11]

The difference in apparent effectiveness between younger and older women suggests that when invasive cancer develops in young women, it is often more rapidly progressive than disease developing in older women, and therefore more difficult to prevent with screening.[15] One possible contribution is the rising incidence of adenocarcinoma. We tested this hypothesis, and found that at all ages the number of unclassified cancers reduced, with a corresponding complementary rise in the diagnoses of adenocarcinoma and other rarer cancer types.[16] Most of the reduction in total cancer cases resulted from reduced rates of squamous carcinoma, so the relative proportion of adenocarcinoma increased. Cytology testing is more effective in the early detection of squamous carcinoma. An important question for human papillomavirus (HPV) testing is whether it will enable the better detection of adenocarcinoma and other types, and thereby reduce these components of the disease as well.

A screening policy must balance the benefits of screening against the potential harms. In Canada, for women under the age of 30 years, around 10% have abnormal smears,[6] thus triggering some form of follow-up. Many are colposcoped, with a substantial proportion being treated by loop electrosurgical excision procedure (LEEP). There is concern that such procedures in young women cause pregnancy-related harm,[17] in addition to the anxiety and disturbance created by informing young women that they have a pre-cancer. The rate of abnormalities found at screening reduces with age, although the likelihood of cancer increases. Thus the balance between benefits and harms changes with age, although it is difficult to identify a specific age when the probability tips.

These data were used to support the Canadian task force recommendations (Box 1),[6] which are similar to those used in the UK, and recommend starting screening later than has been suggested in US, Australian, and previous Canadian policies. They strike a balance between trying to reduce the very rare cancers in young women and the harm caused to the 98% of women who would never develop this cancer. The recommendations have again been opposed by the Canadian societies for obstetrician gynaecologists, gynaecological oncologists, and colposcopists, and those who are loathe to withhold screening and subsequent treatment from women aged between 21 and 25 years.[18, 19] The Canadian provinces have not yet had time to integrate the task force recommendations, but some have already reacted negatively to the increase in screening age. Currently screening usually starts at the age of 21 years.[18] Consequently, colposcopists regularly treat young women with ‘high-grade’ lesions, for which there is high diagnostic variation among pathologists.[20] Given the medico-legal environment, pathologists struggle to accurately describe a lesion and clinicians hesitate to observe an abnormality over time. That most of these lesions will regress, or not progress before a much later age, is overshadowed by concern about missing a rapidly developing cancer.

Box 1. Recommendations of the Canadian Task Force on Preventive Health Care 2013[6]

Classified using the GRADE system[21]

GRADE classifies evidence under three categories: strong, moderate, or weak, based on the likelihood that further research would change the estimate of effect.

Then the recommendation made is either strong or weak, and for or against, based on the quality of the supporting evidence, the degree of uncertainty about the balance between desirable and undesirable effects, the degree of uncertainty about values and preferences, and about whether the intervention represents a wise use of resources.

  • For women under the age of 20 years we recommend not routinely screening for cervical cancer
    • (strong recommendation; high-quality evidence)
  • For women aged 20–24 years we recommend not routinely screening for cervical cancer
    • (weak recommendation; moderate-quality evidence)
  • For women aged 25–29 years we recommend routine screening for cervical cancer every 3 years
    • (weak recommendation; moderate-quality evidence)
  • For women aged 30–69 years we recommend routine screening for cervical cancer every 3 years
    • (strong recommendation; high-quality evidence)
  • For women aged 70 years and over who have been adequately screened (i.e. three successive negative Pap tests in the last 10 years), we recommend that routine screening may cease. For women aged 70 years or over who have not been adequately screened we recommend continued screening until three negative test results have been obtained
    • (weak recommendation; low-quality evidence)

Our data, and conclusions about the age to commence screening, are based on the era before HPV immunisation. Policies will need to change further for the cohorts who are immunised; but for individuals or groups who miss immunisation, or in countries that do not embrace it, this information will remain useful in policy decisions. Screening is effective for slowly developing, common disease, where the benefit can outweigh the harms. It cannot prevent rare, rapidly advancing disease. So no screening programme can be 100% effective: there will always be some failures. The policy should minimise these, in balance with minimising the harms to many.

Disclosure of interests

JAD is a member of the Canadian Task Force on Preventive Health Care, and chairman of the Cervical Screening Guidelines writing group. This position is honorary, supported only by the provision of travel and meeting expenses. The views expressed in this article are those of the authors and do not represent those of the Public Health Agency of Canada or the Canadian Task Force on Preventive Health Care. CP receives consulting and speaker fees from Merck and GlaxoSmithKline.

Contribution to authorship

The idea for the article was suggested by the editor of BJOG, and the article was drafted by JAD. Both ABM and CP revised and edited the article. All three authors had worked on the original papers that this commentary summarises.

Details of ethics approval

No ethics approval was needed for this work.


No funding was received for this work.


Data were provided by the Public Health Agency of Canada.

Commentary on ‘When to start cervical screening: epidemiological evidence from Canada’

Usually an increase in scientific understanding provides clinicians with greater insight into optimal patient management. Unfortunately, our growing understanding of the epidemiology and natural history of high-risk human papillomavirus (HPV) infections and consequent lesions has only contributed to international controversy about the best strategies for cervical cancer prevention through screening. One area of dispute, as Dickinson and colleagues point out in this issue of BJOG, is when to begin screening. More data are unlikely to quiet the argument.

Controversy arises because although cytological screening and the identification and eradication of cervical cancer precursors have dramatically reduced the burden of cervical cancer in developed countries, decisions about when to begin screening, how often to screen, what modalities to employ, and when to end screening remain value judgments. Science can inform debates over these questions, but cannot resolve them. Different societies will develop different solutions based on varying assessments of cost, value, benefits, and harms. Also important yet variable is whether screening is organised, with tracking and recall to minimise default, or opportunistic, with individual women and clinicians determining the screening approach. More intensive screening increases sensitivity for pre-invasive lesions, and may be appropriate when the timeliness of the next screening cannot be assumed.

The incidence of cervical cancer in teens and women in their early twenties is only a few per million (Benard VB et al. Obstet Gynecol 2012;120:1117–23), and the prevalence is lower. At the same time, HPV infections and borderline cytological abnormalities are common in young women. Most lesions in young women, including grade–2 cervical intraepithelial neoplasia (CIN2), resolve without treatment (Moscicki AB et al. Obstet Gynecol 2010;116:1373–80). If the risk of an abnormal screen is 5–10%, then for each cancer found, thousands of women experience the stigmatisation of diagnosis with an HPV-related abnormality, and must go through the inconvenience and pain of colposcopy without benefit. Although the impact of treatment for pre-invasive cervical lesions on subsequent pregnancy is controversial (Stock SJ et al. BJOG 2012;119:647–9), the concept that reducing cervical mass predisposes women to preterm labour is biologically plausible, and should be considered before undertaking mass treatment of lesions that usually regress. Many cancers in this age group are not preventable, as has been shown in the UK (Sasieni P et al. BMJ 2009;339:b2968), and in Figure 1 presenting the results from Canada.

In a technological age, we look for technical solutions, but not all cancers can be prevented. We make choices based on risk. Men don't get mammograms, but 50–year-old women do, although both can get breast cancer. With this in mind, comprehensive vaccination against HPV before first sexual exposure, in combination with organised screening incorporating referral to colposcopy and treatment centres of excellence, offers a more rational strategy for cervical cancer prevention than screening young women. Because clinicians in the USA haven't implemented any of these components well, we begin screening earlier and screen often, despite contrary guidelines. This report from Dickinson and colleagues provides additional information and argument that will allow all to consider screening in the context of local values and limits.

Disclosure of interests

I have no financial relationship with any drug, device, or assay manufacturer. I have held offices with and received honoraria from the American Society for Colposcopy and Cervical Pathology.

  • LS Massad

  • Department of Obstetrics and Gynecology, Washington University, St Louis, MO, USA