Early Detection and Diagnosis
Monitoring a national cancer prevention program: Successful changes in cervical cancer screening in the netherlands
The success of screening, an important cancer prevention tool, depends on the quality and efficiency of protocols and guidelines for screening and follow-up. However, even centrally organized screening programs such as the Dutch cervical screening program occasionally show problems in performance. To improve this program, the screening scheme, follow-up, administration and financing protocols and guidelines were thoroughly changed in 1996. This study evaluates the consequences for the performance of the national program. Five-year coverage rate, the proportion of screened women sent to follow-up, follow-up compliance and duration, and the yearly number of Pap smears before and after the changes in 1996 were compared. Five-year coverage increased substantially in the added target age groups (30–34, and 54–60 years); in the old target age group (35–53 years) it remained around 80%. The percentage of screened women sent to follow-up decreased from almost 19–3% per screening round, due to a more restrictive use of the Pap 2 classification, and an evidence-based cessation of follow-up of negative smears without endocervical cells. Follow-up compliance has improved, and the average time until a woman is either referred or rejoins the regular screening schedule, has become shorter. The total number of smears, a strong determinant of screening costs, has decreased by 20% primarily due to the changed follow-up recommendations. In conclusion, the 1996 changes in protocols and guidelines, and their implementation have increased coverage and efficiency, and decreased the screening-induced negative side effects. © 2006 Wiley-Liss, Inc.
Screening is an important cancer prevention tool. In case of cervical, colorectal and breast cancer it is recommended1 and adopted as national programs in various countries.2, 3, 4 Depending on the outcomes of ongoing trials,5, 6, 7 more screening programs (for prostate, lung and ovary cancer) may in the future be offered to the general population. The success of screening depends on the quality and implementation of protocols and guidelines for primary screening and follow-up. Often, the balance between prevented mortality from cancer, and the screening-induced costs and negative side effects (e.g., higher morbidity, unnecessary treatments and raised anxiety due to false-positive results) is fragile.8, 9 Also, people may be screened too frequently, while others are not screened at all.10, 11 Further, new screening tests become available occasionally,12 and these require thorough evaluation before they can be incorporated into a screening program. Screening therefore requires regular monitoring and adaptation in order to keep the program effective and efficient.
Since the 1980s, cervical cancer screening with the Pap smear has been offered to the population in the Netherlands through an organized program. Pap smear is a non-invasive test and aims to prevent cancer deaths by treating pre-invasive and early invasive disease. Like in the UK in the 1980s,13, 14, 15 the evidence gathered in the Netherlands in the early 1990s pointed towards a suboptimally performing program, in terms of both the organization and the efficiency of screening of the target population.16, 17, 18, 19 In 1993, the Ministry of Health called for an immediate inquiry into possible solutions.20 Subsequently, new protocols and guidelines regarding the screening and follow-up schemes, administration and financing were implemented nationally in 1996 (see Appendix).21 The new screening and follow-up schemes were agreed upon in a consensus meeting of the 5 professional groups implementing cervical cancer screening (pathologists, gynaecologists, general practitioners (GP), regional cancer registries and local health authorities) together with epidemiologists and Ministry of Health representatives. Much emphasis was devoted to assuring that the guidelines would be adhered to.22
Seven main implemented solutions can be distinguished. First, since 1996 the program covers the whole country. Financing and coordination are managed centrally. The program is implemented locally through special regional centers operating in a covenant, i.e. as financially binding associations of the 5 professional groups listed earlier. Second, all non-attendees to the first invitation are systematically sent reminders. Third, reimbursement of preventive smears taken outside of the regular screening schedule, which are considered ineffective and may result in unnecessary diagnostic and treatment procedures,19, 23 has been abolished. Fourth, the screening interval was increased from 3 to 5 years in a broader target age group (30–60 instead of 35–53 years).18 Fifth, new pathology guidelines24 aimed to improve the specificity of the Pap smear by downgrading the considerable number of smears with sole morphocytological signs of inflammation and/or presence of specific microorganisms from a borderline (Pap 2) to a negative (Pap 1) smear. Sixth, the new pathology guidelines aimed to limit the maximum duration of follow-up of borderline (Pap 2) and mildly (Pap 3a1) dyskaryotic smears (BMD) before the final referral or normal screening recommendation. Seventh, an amendment from 2002 ceased to advise a repeat Pap smear to negative smears lacking endocervical cells (Ecc-).25, 26
To assess improvement in the Dutch cervical cancer screening program after the completion of the first 5-yearly screening round under the new protocols and guidelines, we will evaluate the impact of these changes on the five-year coverage rate, the proportion of screened women sent to follow-up, follow-up compliance and duration, and the yearly number of smears.
Material and methods
Performance of the new protocols and guidelines is based on program smears from 2003. The exception is the follow-up compliance, which is instead based on program smears from 1999 in order to have available data for a 4-year follow-up period. This performance of the current program is compared with the situation before 1996, which is based on program smears from 1994.
Information on all cervix uteri cytological and histological tests in the Netherlands registered until 31st March 2004 was retrieved from the nation-wide network and registry of histo- and cyto-pathology (“PALGA”). From 1990 onwards all pathology laboratories were linked to this registry.27
PALGA identifies a woman through her birth date and the first 4 letters of the (maiden) name. To correct the occasional false matches in this identification,28 we excluded the 0.5% most common surnames.29 This applies to the results in Tables I, III, IV and consequently VI, and excludes about 30% of women from the PALGA. Age is defined on December 31st of the analyzed year, in agreement with the definition of age at invitation as used in the screening program.
Table I. Five-Year Coverage Rates in the Netherlands before (1994) and after (2003) the Implementation of the New Cervical Cancer Screening Protocols and Guidelines, for Women Aged 30–64, Per Age Group
For all registered cervical smears, PALGA enables recoding of the reason for smear-taking into program screening, spontaneous screening, medical complaints or follow-up. This information is missing for 52 and 10% of the primary smears in 1994 and 2003, respectively. A program smear is defined as any smear in PALGA that was primary and taken in the calendar year (or the first 3 months thereafter) in which the woman was eligible for the program given her birth year.
Five-year coverage rates are calculated by comparing the number of women who had at least one smear taken for any reason (as counted from PALGA), with the estimated number of women at risk (i.e., with a cervix) alive on January 1st of the analyzed year. The number of women at risk were estimated with the data obtained from the Dutch Central Bureau of Statistics,30 decreased by the estimated number of women with their cervix removed by a hysterectomy.31 Further, we excluded the 0.5% common surnames both in the numerator and the denominator. Because the latter cannot be obtained from any official statistics, we approximated the fraction of women remaining in the population at risk after the exclusion of 0.5% most common surnames by observing the fraction of tests remaining in PALGA after the same exclusion. The implicit assumption here is that the screening behavior and the commonness of surnames are not associated.
A screening episode is defined as starting with a primary test and eventually followed by secondary tests (smears or biopsies). Follow-up or secondary tests are tests made within 4 years of an abnormal smear, an inadequate quality smear or a non-negative biopsy after which follow-up has not yet been completed according to the guidelines. All other tests are seen as primary tests starting the episode. An episode is finished either when follow-up is complete according to the current guidelines (e.g., 2 consecutive negative smears after a BMD smear, or 3 consecutive negative smears after treatment of histologically confirmed cervical intraepithelial neoplasia), or when there are no more tests registered in 4 years. The categorization of smear results and the corresponding follow-up recommendations are presented in the Appendix.
Excess smears are defined as all smears that are taken in a certain period that do not contribute to the observed coverage in the target population. These may be due to e.g. more frequent screening than it is recommended, but also due to secondary (diagnostic) testing. The number of excess smears is presented for the relevant screening interval and 1,000 women per year; it is equivalent to (total yearly number of smears – number of smears needed yearly to reach the observed coverage) × 1,000/number of women in the target group. The number of smears needed yearly to reach the observed coverage equals (the population at risk in the target age range × observed coverage) / recommended interval.
Five-year coverage in the target population reached 77% in 2003 (Table I), which is 12% points higher than the response to screening invitations, i.e. the attendance rate of 65% (data not shown). This difference is due to smear-taking outside the program. The coverage in 1994 in the then targeted age group (35–53 years) was higher than in the newly targeted age group in 2003 (30–60 years). However, given that based on evidence it has been decided since 1996 that a larger age group is to be reached, the situation has improved. The coverage increased substantially in the added target age groups, while in the age groups targeted both before and after 1996 it remained at about the same level.
One percent of the program smears was of inadequate quality for evaluation in 2003 (Table II). Moderately dyskaryotic or worse abnormalities (>BMD) were found in 0.5% of program participants, and further 1.8% of women had a BMD smear. The remaining women, 96.7%, had a negative smear. Compared with the situation in 1994, there are two interesting differences. First, there was a >80% decrease in the frequency of a follow-up smear advice. This was caused by the change in the definitions regarding the classification of sole morphological signs of inflammation. Second, immediate referral recommendation because of high-grade cytological abnormalities has almost doubled in frequency because since 1996 moderately dyskaryotic smears (Pap 3a2) also fall into this category (see Appendix).
Table II. Primary Smear Results in the Organized Cervical Cancer Screening Program Before (1994) and After (2003) the Implementation of the New Protocols and Guidelines, by Type of Follow-up Recommendation
| Due to a negative smear||84.8%||81.1%|
| Due to an Ecc- smear1||11.9%||7.3%|
| Follow-up smear||1.8%||10.0%|
| Immediate referral||0.5%||0.3%|
| Repeat smear due to inadequate quality||1.0%||1.3%|
Among women with a 6-month follow-up smear advice (BMD smears), 74% received follow-up within 9 months, which we define as timely follow-up according to the guidelines (Table III). Compared with 1994, the compliance with follow-up guidelines has increased substantially. There has also been an increase in longer-term follow-up (measured within 4 years). The time in follow-up for these women until the final referral or a screening recommendation was shorter than in 1994 (Table IV). This difference is especially notable for women with a negative follow-up outcome: they were on average recommended to rejoin the normal screening schedule 12 months earlier than in 1994. Women with a positive follow-up outcome were on average referred to a gynaecologist 4 months earlier. Moreover, after on average 4.6 years, fewer women had no clear final follow-up recommendation in 1999 than in 1994. The average number of tests performed before the final either referral or a screening recommendation decreased by 10% (data not shown).
Table III. Compliance with follow-up After Primary Screening Program Smears Before (1994) and After (1999) the Implementation of the New Protocols and Guidelines, by type of follow-up Recommendation
|Percentage of program smears||2.1%||10.0%|
| of which with timely follow-up1||74%2||44%2|
| of which followed-up in 4 years*||90%||78%|
| of which with a biopsy (of*)||27%||14%|
|Percentage of program smears||0.6%||0.3%|
| of which with timely follow-up1||85%||91%|
| of which followed-up in 4 years*||97%||97%|
| of which with a biopsy (of*)||94%||95%|
|Repeat smear due to inadequate quality|
|Percentage of program smears||0.8%||1.3%|
| of which with timely follow-up1||42%||45%|
| of which followed-up in 4 years*||85%||85%|
| of which with a biopsy (of*)||3%||18%|
Table IV. Final Recommendation for women with a 6-month Follow-up Smear Advice (To Rejoin the Normal Screening Schedule, or a Referral to Gynaecologist), and Average Duration Until Final Recommendation, Before (1994) and After (1999) the Implementation of the New Protocols and Guidelines
|Sent back to screening||32%||24%||25||37|
|Without any follow-up||9%||16%||n.a.||n.a.|
|Total||100%||100%|| || |
In the recent period, 85% of the women with a direct referral advice (>BMD smears) received follow-up in time (3 months), which increased to 97% within 4 years (Table III). This compliance has not improved since 1994, but did also not worsen even though the new protocols on immediate referral include the moderately dyskaryotic smears (Pap 3a2). Among the 97% of followed-up women, 94% had at least one biopsy. For smears that need to be repeated due to inadequate quality the follow-up compliance has remained similar. However, contrary to recommendations these women were in 1994 more often referred and biopsied immediately instead of having a repeat smear taken.
In 2003, program smears represented 67% of all smears, and those taken for medical complaints 13%. Spontaneous screening was small in volume at <2% of the total screening activity. Together with those smears for which no reason for smear-taking is given, primary smears add up to 91% of all smears (Table V). Secondary smears account for 9% of all smears.
Table V. Numbers of Primary and Secondary Smears (× 1,000), by Screening Age Group, Before (1994, Adjusted for Population Growth Between 1994 and 2003) and after (2003) the Implementation of the New Protocols and Guidelines
|<30||28||106||4||26||32 (4%)||132 (13%)|
|30–64||666||598||61||214||727 (92%)||812 (83%)|
|>64||26||29||3||9||29 (4%)||37 (4%)|
|Total||720 (91%)||733 (75%)||68 (9%)||248 (25%)||788 (100%)||981 (100%)|
In total, 787,506 smears were taken in 2003, which represents an overall drop of 20% since 1994 (Table V). Compared with 1994 and adjusted for population growth, the number of primary smears was about 2% lower in 2003. The decreased number of primary smears coupled with an increased coverage in a broad age range must have been associated with a longer average interval between consecutive primary smears. Indeed, both the frequency of very short screening intervals, as well as of the intervals of 2–4 years decreased (data not shown). The effect of the screening scheme protocol change is here intertwined with better adherence to these protocols. Moreover, there was less screening outside of the target ages, especially among women before their thirties (Table VI). Secondary smears dropped by 73% between 1994 and 2003 (Table V). This drop was in approximately two-thirds due to how smears are evaluated with regard to inflammation (i.e., the Pap 2 definition change), and in the remaining third due to cessation of follow-up to Ecc- smears since January 2002.
Table VI. Summary: Effectiveness, Screening-Induced Side Effects and Efficiency Indicators Before and After the Implementation of the New Protocols and Guidelines in 1996
| Coverage in the target age group1||77%||82%|
| Overall compliance to follow-up2||90%||79%|
|Screening-induced negative side effects|
| Proportion of women with a follow-up advice||3%||19%|
| Average time in follow-up before sent back to screening (women with a follow-up smear advice)||25 months||37 months|
| Number of excess smears per year per 1,000 women in the target age group3||76||261|
| Number of smears per 1,000 women aged 20–29, per year||33||111|
Since the 1996 changes, the performance of the Dutch cervical cancer screening program has improved considerably. With the same number of 7 smears per lifetime in a broader age range, the 5-year coverage in the added target age groups (30–34, and 54–60 years) rose substantially to 70% or above, with a loss of a few percent in the old target age group (35–53 years) where the coverage remained around 80%. This better coverage was achieved with a 20% lower number of smears. The share of women sent to anxiety-causing follow-up decreased from 19 to 3% per screening round. Ninety percent of these women, an increase of 11% points, were followed-up after their non-negative screening smear. They spent a considerably shorter time in follow-up before they were either referred, or rejoined the regular screening schedule (Table VI). These results contribute towards gains in effectiveness and efficiency, and a reduction in screening-induced negative side effects and costs associated with cervical cancer screening in the Netherlands. Because before 1996, for slightly more than a half of the primary smears, the reason for smear-taking was not known, in an additional analysis we included these smears. The pooled estimate of smear non-negativity rate in 1994 was 16% instead of 19% reported in Table II, and the recent 3% still represents an 81% reduction since 1996 (instead of 84% reduction compared to 19%).
Coverage is a major effectiveness determinant of cervical cancer screening.32, 33 In the age groups targeted both before and after 1996 (35–53 years) a small decrease in 5-year coverage rates was observed (Table I). It appears that the lower frequency of invitations (one invitation every five instead of 3 years) has been counteracted by the beneficial effects of the 100% (i.e., nation-wide) invitation coverage and systematic reminders. Recent evidence34 has shown that the uptake of screening is higher when the invitation is sent by the GP as opposed to the municipality. Further expansion of a GP invitational system is currently under review. The English experience with introducing graded financial incentives for the GPs to stimulate coverage has been positive,35 though feasibility and acceptability of implementing such an instrument in the Dutch health care is not known.
Sending fewer women to follow-up, i.e. an increase in test specificity, has the potential of decreasing test sensitivity. Because of the low predictive value for (future) cancer of borderline dyskaryotic smears,36, 37 and the long average screen-detectable preinvasive period for cervical cancer,38, 39, 40 it is expected that when women are regularly screened, less intensive follow-up will only lead to a very small loss in program sensitivity. The CIN 3+ detection rates, i.e. the number of histologically confirmed cervical intraepithelial lesions grade 3 or higher per 1,000 women screened, increased from 4.4 to 5.4 (i.e., by 22%) between 1994 and 1999 (corrected for the extent of default to follow-up; data not shown). If anything, this does not point at lower sensitivity. However, even assuming no change in sensitivity, this increase was expected due to both an earlier start of program screening (higher incidence and prevalence of CIN in young women), and a longer screening interval in 1999 than in 1994 (higher within-interval cumulative incidence of CIN 3). It is unknown to what extent changes in the detection rates are due to changes in the degree of over-diagnosis. A better indicator of the potential change in program sensitivity, i.e. the interval cancer rate, is going to be examined in later analyses.
There is quite a diversity in how cervical cancer screening is organized in Europe,3, 41 and its performance varies widely.42 Compared to other countries with organized national programs (Table VII), the Netherlands has been successful in limiting the number of excess smears while maintaining a high coverage rate. For example, even though the recommended screening interval is 5 years in Finland, the yearly number of smears is enough to screen each woman in the target age every 2.1 years, compared to every 4.7 years in the Netherlands. There are several reasons for this phenomenon. Contrary to Finland, procedures exist in the Netherlands (likewise in Sweden and England) that help sort out women with very recent smears. Next, smears taken outside of the regular screening schedule are in the Netherlands only reimbursed when the woman has medical complaints. Finally, Pap-smear screening in the Netherlands primarily takes place in GP practices, and much attention has been paid in the Dutch guidelines for GP practices to the frequency of smear-taking.49
Table VII. Comparison of the Current Dutch Cervical Cancer Screening Program with Other Cervical Cancer Screening Programs in Terms of Process Indicators42, 43, 44, 45, 46, 47, 48
|Cervical cancer mortality per 100.0001||1.9||0.9||2.1||1.7|
|Organized program characteristics|
| Target age group||30–60||30–60||25–64||20–60|
| Screening interval (years)||5||5||3–5||2–4|
| Number of recommended smears per lifetime||7||7||12||14|
|Organized program performance|
| Share of smears within the program2||67%||37%||75%||39%|
| Target population / total number of smears||4.7||2.1||3.5||2.5|
| 5-year coverage in the target group||77%||93%||81%||<82%3|
| % Inadequate quality||1.0%||0.01%||9.3%||n.a.|
| % Positive smears4||2.3%||7.3%||6.4%||1.5%|
| Of which highly positive smears||0.5%||0.6%||1.1%||n.a.|
The structure of follow-up advices given in the program also differs among countries. While the Netherlands and Finland both have a low rate of inadequate smears, the sample inadequacy rate in England is close to 10%. This is one of the reasons why liquid-based cytology is considered a cost-effective option in England50 and not in the Netherlands.51 England and Finland have a high positivity rate of around 7%. The Netherlands and Sweden are both low with 2.3 and 1.5%, respectively. The fact that the bulk of the differences disappears if one compares the frequency of only highly positive program smears is consistent with the view that the countries are using widely differing definitions of BMD abnormality. Therefore, the planned comparison of interval cancer rates before29 and after the sharp decline in the proportion of BMD smears in the Netherlands will also be interesting from an international perspective.
The observed improvement in the Dutch cervical cancer screening program was achieved in a short time. Implementation of protocols and guidelines is a complex process that depends on both the protocols and the guidelines themselves as well as on the context in which they are implemented.52, 53 In the Netherlands, the new protocols and guidelines24, 49, 54 (see Appendix) were developed by recognized organizations of those professionals who perform smear-taking and pathology evaluation, and were based on available evidence. In many instances, such as for follow-up, they simplified and made clearer the recommended algorithms of care. In turn, adherence to the new protocols and guidelines was stimulated with those instruments that have proved to be effective in prevention, e.g. educational outreach visits and specialized software modules.22, 55 Responsibility for implementing the program according to the agreed protocols, which was also secured with financial stipulations, was put in the hands of the regional screening centers that are backed by the covenant of the 5 professional groups with an interest in cervical cancer prevention. Early encouraging evidence from the new program56 could have further helped stimulate adherence to the pace set out in the protocols.
In conclusion, the 1996 evidence-based change of protocols and guidelines regarding the screening scheme, follow-up, administration and financing has brought about a considerable improvement in the cervical cancer screening program in the Netherlands. It was made possible by continuous evaluation based on national pathology registry data, timely dissemination of results and carefully developed change and implementation phases.
The authors thank Statfin and the Swedish National Board of Health and Welfare for providing the population data for Finland and Sweden, respectively.
Appendix: Appendix – Overview of the changes in the cervical cancer screening protocols and guidelines in the Netherlands in 1996
Table AI. Protocols and Guidelines Regarding Screening, Financing and Administration in the Cervical Cancer Screening Program in the Netherlands, Before and After 1996
Table AII. Guidelines Regarding Follow-Up in the Cervical Cancer Screening Program in the Netherlands, Before and After 1996