Abstract
Material and methods. Based on a five-year follow-up of 130 692 women who participated in the five-yearly organised screening in Finland in 1995, we estimated the risk for incident invasive cervical cancer and severe precancerous lesion for groups of women with different combinations of high-risk determinants at baseline. Results. Compared to women with normal cytology and no symptoms at the index screen, the risk of developing invasive cervical cancer during the next screening interval was significantly elevated in women with abnormal cytology at baseline screen (RR 5.8, 95% CI 2.6–13.1); additionally, the risk of cervical intraepithelial neoplasia grade 3 (CIN3) lesions was even higher (RR 9.4, 95% CI 6.9–12.5), both results after removing respective screen detected cases at index screen. Among women with symptoms of vaginal bleeding or abnormal vaginal discharge but normal cytology the risk of invasive cervical cancer was increased but not significantly (RR 2.4, 95% CI 0.9–6.3) and the risk of CIN3 lesions was only slightly elevated (RR 1.5, 95% CI 0.95–2.4). In women less than 40 years of age, the risk of cervical cancer was 50% higher than in women 40 years or more in those with normal cytology and no symptoms, however not significantly. Symptoms indicated a similarly elevated risk for cancer in both age groups. After abnormal cytology there was a nine-fold risk for cancer in young women (under 40 years) and a six-fold risk in old women (40 years or more) when compared to women aged 40 or more with no symptoms and normal cytology. In young women CIN3 lesions were found in great excess, three times more often, compared to those aged 40 and over. Discussion. In all, selective intensified screening for cervical cancer complementing the five-yearly unselective programme is feasible for women with abnormal cytological findings at an acceptable level of incremental cost.
An organised programme for population-based cervical cancer screening with the conventional Papanicolaou smear test was introduced in Finland in 1963 by the Cancer Society of Finland (CSF) [Citation1]. Since the beginning of the 1990s the basis of the Finnish routine cervical cancer screening programme has remained virtually unchanged: in most municipalities all resident women from 30 to 60 years of age are offered free cytological screening once every fifth year.
In the 1960s, there was no evidence on population-level effectiveness of organised cervical cancer screening programmes. Because of this, selective screening – i.e. offering free cervical screening to women with specific high-risk determinants – was seriously considered as an option as it would have been a less costly option compared to unselective population-based screening. Information on age, sexual behaviour and fertility were suggested to identify groups of low and high risk of cervical cancer; in addition, characteristics of cytological diagnosis as well as various symptoms and gynaecological infections were suggested as potential high-risk determinants [Citation2,Citation3].
Although cervical cancer screening was launched in Finland as an unselected population based programme, detailed individual level information was collected at the screening visit. This information was later used to evaluate the selective screening approach. Based on the information collected at screening visits and supported by information from similar studies on breast cancer screening, it was shown that unselective screening was more beneficial at population level than selective screening [Citation4,Citation5]. The results also showed that the severity of cytological diagnosis was directly related to detection of severe dysplasia or cervical cancer. Vaginal infections seemed to be associated with a high cervical cancer risk during follow-up, as well as coital bleeding [Citation6,Citation7]. Based on these observations it seemed plausible that combining selective screening to the population-based programme – i.e. inviting the women with high-risk determinants not only five-yearly but also between these five-yearly visits – would further increase the programme effectiveness.
Consequently, intensified screening for defined high-risk groups was gradually introduced into the Finnish screening programme. Since late 1980s this selective (also called “risk group”) screening practice has covered the women with borderline or mild cytological changes (Papanicolaou class II equal to reactive changes or ASC-US in Bethesda 2001 and borderline changes in British terminology) found at index screen. Also women with more severe cytological changes (Papanicolaou class III-V; ASC-H, LSIL or worse; mild dysplasia or worse) but with no histological confirmation of cervical intraepithelial neoplasia (CIN) or carcinoma found in colposcopically directed biopsies, and women who reported symptoms of non-menstrual bleeding or spotting at the screening visit have been included in the risk group. These women are generally re-invited to organised screening within 12–24 months.
The aim of this register-based study was to compare intensified selective screening based on symptoms and findings at the index screening visit (so called risk group screening) with unselective population-based screening in Finland by assessing risk estimates during the next screening interval for cervical cancer and its severe precursors (cervical intraepithelial neoplasia grade 3, CIN 3) for groups with different high-risk determinants (abnormal cytology (Papanicolau class II or more severe with no invasive treatment given), and/or symptoms of vaginal bleeding and abnormal vaginal discharge) at baseline.
Material and methods
The baseline population comprised of 130 895 women aged 20 to 65 years in 1995 who attended the five-yearly cervical cancer screening within the population-based programme. Follow-up started six months after the index screening visit in 1995 and it ended either six months after the subsequent five-yearly screening visit in 2000, at the time a lesion of interest (invasive cervical cancer, ICC or cervical intraepithelial neoplasia grade 3, CIN3, including dysplasia gravis, epidermoid ca in situ, and adenocarcinoma in situ) was diagnosed, at date of death, or date of emigration, whichever came first. When there was no data on the subsequent screening visit in year 2000 (n=37 774) due to non-compliance, missing invitation or reaching the upper age limit, the follow-up ended by December 31, 2000 at the latest. Women diagnosed with invasive cervical cancer within six months from the index screen in 1995 (n=22) were assumed to have screen-detected lesions, and they did not contribute to the follow-up time (excluded from analyses). Concerning analyses with CIN3 as the endpoint, also women with a CIN3 lesion (n=123) diagnosed within six months from the index screen were excluded (as index screen-detected findings). Likewise, invasive cervical cancers and CIN3 lesions diagnosed within six months from the subsequent five-yearly screen in 2000 were defined as screen-detected.
The women in the baseline population already had a screening history from one to nine screening rounds. The women were followed for five years to the next screening round as specified above. The data on screening invitations, participation and findings within the organised screening programme were derived from the Mass Screening Registry (MSR) of the Finnish Cancer Registry (FCR). The MSR data was compiled by screening laboratories. They followed actively all women who were referred for colposcopy to obtain data on final diagnosis and histology. The diagnosis of colposcopy including data on histology was then sent to the MSR along with data on screening findings. The baseline data was then linked at individual level to the nation-wide cancer database (FCR) for malignant and pre-malignant primary tumours of cervix uteri, and to the Central Population Registry for information on emigration and deaths. Using the additional linkage data we excluded women who had been previously diagnosed with invasive cervical cancer (n=50), women who moved out of the country (n=33), or women who died within six months (n=74) from the index screening visit, i.e. prior to start of follow-up; additionally the few (n=12) passive personal identification numbers included in the baseline data were excluded. After restricting the data to those women who had their index screen visit between January 1 and December 31, 1995, a total of 130 692 women were included in the final cohort.
The final data included individual level information on organised screening invitations (five year interval invitations and intensified risk-group invitations based on the findings or symptoms at the screen), screening visits (including information on various baseline symptoms), cytology results (registered as Papanicolaou class) and colposcopy referrals with clinical or histological diagnoses within the organised programme, and invasive cervical cancers (ICC) and severe pre-invasive lesions (CIN3) detected outside the organised screening programme. Data on ICC and CIN3 lesions were primarily derived from the Finnish Cancer Registry database. Concerning invasive cervical cancers this database is practically complete [Citation8], but for pre-invasive lesions some lesions may be missing. Thus, concerning information on CIN3 lesions, also data from the screening visits were used among attendees.
In the final cohort we studied the risk of being diagnosed with invasive cervical cancer (ICC) and with severe pre-invasive cervical lesion (CIN3) between screens taking into account the individual status of self-reported symptoms and cytological findings at the index screening visit. Relative risk (RR) estimates for groups of women with different index screen statuses were calculated using Poisson regression with follow-up time as a continuous variable. The group of women with normal cytology and no self reported symptoms at the index screen was used as the reference group. In addition to crude risk estimates with 95% confidence intervals (95% CI), age-adjusted estimates were calculated with age as a categorical variable. To study the effect of age on the risk estimates, we repeated the Poisson regression analyses with combined categories including index screen status and age categorised into two groups: less than 40 years of age and those aged 40 or more. The reference category was the one with the lowest risk, women aged 40 years or more with normal cytology and no symptoms at index screen.
Results
Of the 130 692 women screened in 1995, 22 were diagnosed with an invasive cervical cancer at baseline and were excluded from follow-up. Among the remaining 130 670 women we observed, during an average follow-up period of 5.0 years (59.6 months) 30 incident invasive cervical cancers between the two five-yearly screening visits and five at the subsequent screen. In addition, we observed 141 cases of incident CIN3 lesions between five-yearly screens and 87 at the subsequent screen. Thus, altogether 263 women (0.2%) were diagnosed with a CIN3 lesion or worse during the entire follow-up ().
Table I. Baseline status at the index screen in 1995 and detection of invasive cervical cancers (ICC) and severe precancerous lesions (cervical intraepithelial neoplasia grade 3, CIN3) during five years among those invited for organised five-yearly screening in Finland.
Of the 130 670 women with no invasive cervical cancer at baseline 17 941 (13.7%) had some symptoms or abnormal cytology or both; 9 992 women (7.6%) had symptoms but normal cytology, 6 982 women (5.3%) had no symptoms but abnormal cytology, and 988 (0.8%) women had both symptoms and abnormal cytology at the index screen visit (). A total of 17 562 women fulfilled the current criteria for Finnish selective screening practice, i.e. they had abnormal cytology and/or symptoms at baseline but no CIN lesions to be surgically treated (loop or excision) during the index screening round. Of these women, 11 826 were invited for intensified screening at least once between 1996 and 1999, and 8 782 women (74% of those invited and 50% of those fulfilling the criteria) attended at least once.
Of the 35 women diagnosed with invasive cervical cancer during follow-up, 22 had normal cytology and no symptoms at the index screening visit, whereas eight had abnormal cytology, and five had symptoms of vaginal bleeding or abnormal vaginal discharge. Of the 228 CIN3 lesions detected during follow-up 127 were found in women with normal cytology and no symptoms at the index screen, 73 had abnormal cytology, 21 had symptoms of vaginal bleeding or abnormal vaginal discharge but normal cytology, and seven had both abnormal cytology and symptoms.
The risk of developing invasive cervical cancer during follow-up (next five yearly screening included) was significantly elevated in women with abnormal cytology at baseline (age-adjusted RR 5.8, 95% CI 2.6–13.1) () compared to women with normal cytology and no symptoms at the index screen. In women with vaginal bleeding or abnormal vaginal discharge but normal cytology, the relative risk estimates were somewhat elevated (RR 2.4, 95% CI 0.9–6.3) but the cases were few and the results not statistically significant.
Table II. Crude and age-adjusted relative risk (RR) of invasive cervical cancer (ICC) and severe precancerous lesion (cervical intraepithelial neoplasia grade 3, CIN 3) among women with any symptoms and/or abnormal cytology at the index screen in 1995 compared to those with normal cytology and no symptoms, reported with 95% confidence intervals (CI).
At the level of CIN3 lesions, the relative risk estimates were similar to those at the level of invasive cervical cancer: the risk of developing a CIN3 lesion was significantly elevated among women with abnormal cytology at baseline (age-adjusted RR 9.4, 95% CI 6.9–12.5), as well as among the women with abnormal cytology at the index screen in combination with vaginal bleeding symptoms or abnormal vaginal discharge (age-adjusted RR 5.6, 95% CI 2.6–12.1). The women who reported symptoms of vaginal bleeding or abnormal vaginal discharge but had normal cytology had a slightly elevated risk of CIN3 (RR 1.5, 95% CI 0.95–2.4) compared to women with no symptoms and normal cytology at baseline ().
Age had a significant modifying effect on the risk estimates. Those with abnormal cytology had a nine-fold risk for cervical cancer in young women (<40 years) and six-fold in old women (40 years or more) compared to the group of old women with normal cytology and no symptoms at index screen (). The risk of CIN3 lesions was almost 40-fold in young women and ten-fold in old women with abnormal cytology compared to the old women with normal cytology and no symptoms (). When young women (less than 40 years) are compared to women aged 40 years or more, the ratio of relative risks is more than three-fold for CIN3 lesions and 1.6-fold for invasive cancers.
Table III. Relative risk (RR) of invasive cervical cancer (ICC) or severe intraepithelial neoplasm (cervical intraepithelial neoplasia grade 3, CIN3) by index screen status among women more than 40 years of age with normal cytology and no symptoms as the reference group, reported with 95% confidence intervals (CI).
Discussion
High grade cytological abnormalities imply an increased risk of subsequent cervical cancer [Citation7,Citation9–11] and thus, women with such findings are referred for colposcopy and treated if a high grade (CIN2/CIN3) lesion is found in biopsies. The proper management of women with bleeding symptoms or abnormal vaginal discharge and those with borderline or mild cytological abnormalities is less clear – should these women be offered reassurance and be returned to the five-yearly screening programme or, instead, should they be alerted and offered intensified screening? In the current study we observed, during a follow-up of maximally five years (next screen detected cases included), an approximately six-fold risk of invasive cervical cancer and a nine-fold risk of CIN3 lesions among women with abnormal cytology at baseline compared to women with normal cytology and no symptoms. However, among women with text book symptoms of cervical cancer, i.e. abnormal vaginal bleeding and/or discharge but normal cytology, the risk of invasive cervical cancer or CIN3 was not significantly increased. In addition, in women less than 40 years of age the risk of an incident CIN3 lesion was five-fold in contrast to women aged 40 or more, even when cytology was normal and no symptoms were present. Approximately 40% of the incident cervical lesions during the five-yearly follow-up period occurred in the specific sub-population, which can be targeted by selective screening. Selective screening based on abnormal symptoms when cytology is normal resulted in a minor proportion of the lesions observed during follow-up.
Cost-effectiveness of an organised screening programme is known to be much higher than that of opportunistic (spontaneous) screening [Citation12]. The additional costs of complementary selective screening are likely to be marginal, given that this additional activity is performed in an organised way, optimally as an extension of a well-operating population-based programme. In this case the resources for health services can be best used and many of the screening-related harms avoided. However, in a population where organised screening has been ongoing successfully for several decades, invasive cervical cancer is a rare disease – especially among the women who participate in organised screening – for which the cost-effectiveness of intensified selective screening complementing the organised five-yearly programme may be low.
In the current study there are some obvious limitations. First, the information on endpoints was primarily based on the Finnish Cancer Registry in which reporting of CIN3 lesions is incomplete. For this, estimates for CIN3 are possibly lower than they would be with complete data, despite the fact that we used the screening database to compensate for underreporting. On the other hand, diagnosis of pre-cancerous lesions, even at the level of CIN3, always include some clinical over-management (i.e. over-diagnosis) because some of the pre-cancers would never progress to invasive cancer and the progression potential of a CIN3 lesion cannot be predicted. These limitations do not affect the risk estimates of invasive disease. Nevertheless, at the level of invasive cervical cancer the most important limitation is the low incidence of invasive cancer in the screened population: even with the 650 000 woman-years of follow-up we did not reach significant risk estimates for sub-groups with symptoms but normal cytology presenting a moderately increased risk. Further, cervical cancer screening in Finland is currently both organised and opportunistic, and it has been estimated that spontaneous cervical smears are taken twice the number of organised smears [Citation12]. The Finnish Cancer registry database does not have information on how the cancer or pre-cancer was detected – i.e. we cannot directly differentiate the lesions diagnosed due to spontaneous screening or normal clinical activity between the five-yearly screenings. More importantly, we do not know how many of the lesions diagnosed between the five-yearly screenings would have been detected at the subsequent screen in the absence of spontaneous screening and the current intensified screening practice.
Additional issues to be taken into account were that the symptoms of vaginal bleeding and abnormal vaginal discharge were self-reported. This causes some uncertainty as women may misunderstand the meaning of “abnormal” and report basically normal conditions as abnormal symptoms; on the other hand, some women may not report even severe symptoms. In theory, selective screening based on self-reported symptoms is also subject to intentional misuse of the free service: reporting abnormal symptoms at the screening visit allows for free annual testing. Hormone replacement is rather widely used among postmenopausal women in Finland, and many of the pre-menopausal women use preventive pills; these exogenous hormones quite commonly cause spotting, which may be falsely interpreted as abnormal bleeding. Irrespective of the reason, it seems that nowadays the symptoms of abnormal vaginal bleeding and discharge are not as clearly associated with increased cervical cancer risk as they used to be in the past. From this perspective it is probably more appropriate to base selective screening on measurable risk determinants, such as presence and severity of cytological abnormalities, rather than on self-reported symptoms.
As long as representing over-diagnosis, CIN3 lesions are an indication of potential harm, and invasive interval cancers are an indication of failure of screening. In the current study with a five-year follow-up we observed a significant six-fold risk of incident invasive cervical cancer and a nine-fold risk of CIN3 lesions among women with borderline or mild cytological abnormality at index screen compared to women with normal cytology and no symptoms. Among women with abnormal symptoms but normal cytology, the risk of invasive cancer or a CIN3 lesion was only marginally increased. Further, the risk of a CIN3 lesion, but not of invasive cancer, was significantly increased among women with normal cytology and no symptoms and aged less than 40 years compared to women aged 40 or more. In women with abnormal cytology at baseline a similar difference by age was observed. Therefore, we recommend intensified screening in women with abnormal cytology. When performed as an extension of a population-based routine screening programme, selective intensified screening in those with abnormal cytological findings is a feasible means to improve programme sensitivity with a minor incremental cost.
Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
References
- Hakama M, Joutsenlahti U, Virtanen A, Rasanen-Virtanen U. Mass screenings for cervical cancer in Finland 1963–71. Organization, extent, and epidemiological implications. Ann Clin Res 1975;7:101–11.
- Clemmesen J. Statistical studies in the aetiology of malignant neoplasms. I. Review and results. I. Acta Pathol Microbiol Scand Suppl 1965;174:1–543.
- Clemmesen J. Statistical studies in the aetiology of malignant neoplasms. II. Basic tables, Denmark 1943–57. II. Acta Pathol Microbiol Scand Suppl 1965;174:1–319.
- Hakama M, Pukkala E, Saastamoinen P. Selective screening: Theory and practice based on high-risk groups of cervical cancer. J Epidemiol Community Health 1979;33:257–61.
- Hakama M, Pukkala E. Selective screening for cervical cancer. Experience of the Finnish mass screening system. Br J Prev Soc Med 1977;31:238–44.
- Viikki M, Pukkala E, Hakama M. Bleeding symptoms and subsequent risk of gynecological and other cancers. Acta Obstet Gynecol Scand 1998;77:564–9.
- Viikki M, Pukkala E, Hakama M. Risk of cervical cancer subsequent to a positive screening cytology: Follow-up study in Finland. Acta Obstet Gynecol Scand 2000;79:576–9.
- Teppo L, Pukkala E, Lehtonen M. Data quality and quality control of a population-based cancer registry. Experience in Finland. Acta Oncol 1994;33:365–9.
- Melnikow J, Nuovo J, Willan AR, Chan BK, Howell LP. Natural history of cervical squamous intraepithelial lesions: A meta-analysis. Obst Gynecol 1998;92:727–35.
- Wright TC, Jr., Cox JT, Massad LS, Twiggs LB, Wilkinson EJ. 2001 Consensus guidelines for the management of women with cervical cytological abnormalities. JAMA 2002;287:2120–9.
- International Agency for Research on Cancer. IARC Handbooks of Cancer Prevention Volume 10: Cervix Cancer Screening. Lyon: IARC Press; 2005.
- Nieminen P, Kallio M, Anttila A, Hakama M. Organised vs. spontaneous Pap-smear screening for cervical cancer: A casecontrol study. Int J Cancer 1999;83:55–8.